Judge Thomas Penfield Jackson bases his rul-ing against Microsoft on the claim that the com-pany’s monopoly in operating systems is protect-ed by an “applications barrier to entry” made upof 70,000 Windows-based software programs.

Without an entry barrier, any dominant pro-ducer that seeks to restrict sales in order to raiseprices above competitive levels will find its mar-ket share eroded as new entrants capture price-sensitive customers. But, according to JudgeJackson, to enter the operating-system market anewcomer would need a large and varied base ofcompatible applications like those available toconsumers who might otherwise chooseWindows. He concludes that “the amount itwould cost an operating system vendor to create[70,000] applications is prohibitively large.”

Judge Jackson seems unaware that the mereexistence of a large number of Windows-basedapplications proves that Microsoft has stirredcompetition among software developers—lead-ing to better products and falling prices and rais-

ing the value of both hardware and software toconsumers. That said, there is a fatal flaw in thejudge’s argument: The overwhelming majority ofthe 70,000 Windows applications that make upthe supposedly impregnable barrier to entryeither never existed as unique products, nolonger exist, or are totally out of date. When onlyunique Windows applications are counted—set-ting aside various versions of the same pro-gram—the number of applications is a small frac-tion of the judge’s count.

Moreover, survey data indicate that theneeds of active computer users are satisfied bya very small number of applications. Thatmeans the barrier to entry into the operating-system market is nowhere near as impregnableas the judge has claimed, which in turn helpsexplain many of Microsoft’s aggressive busi-ness tactics to preserve its market position.Because the judge’s most essential finding isclearly erroneous, it cannot support his conclu-sions of law.

Microsoft’s “Applications Barrier to Entry”The Missing 70,000 Programs

by Richard McKenzie

_____________________________________________________________________________________________________

Richard McKenzie is a professor in the Graduate School of Management at the University of California, Irvine;author of Trust on Trial: How the Microsoft Case Is Reframing the Rules of Competition (Boston:Perseus, 2000); and an adjunct scholar of the Cato Institute.

Executive Summary

No. 380 August 31, 2000

Introduction

Judge Thomas Penfield Jackson, whopresided over the Microsoft antitrust trial,understands that, in order for a monopoly tosuccessfully restrict sales with the intent ofraising prices and profits, there must besome barrier that keeps competitors out ofthe market. Without an entry barrier, higher-than-competitive prices and profits wouldattract rivals that could enter the market tocover sales not made by the would-bemonopolist.

The judge recognizes that there are nolegal barriers to entry into the operating-sys-tem business, such as those that protect theU.S. Postal Service from rivals who mightwant to deliver first-class mail. There are alsono critically important resources in softwaremarkets—unlike, say, the diamond market—that could be cornered and denied to rivals.After all, the core resource of operating sys-tems is a sequence of 1s and 0s, and suchsequences can be produced ad infinitum toform alternative operating systems all overthe world.

In lieu of standard barriers to entry, JudgeJackson has fashioned a totally new entry bar-rier that has never before been used inantitrust cases, dubbed the “applications bar-rier to entry.”1 If for no other reason, that newentry barrier needs to be carefully scrutinizedbecause it may not, on close examination, bewhat the judge says it is (i.e., the kind ofmonopoly protection that would warrant thebreakup of the company and the bindingbehavioral restrictions the judge would like tosee imposed on Microsoft).

The Existence ofCommercially Viable

AlternativesAny barrier to entry would be important

in the operating-system market becauseJudge Jackson found that there is “no com-mercially viable alternative” to Windows,2

which implies that any potential competitorwould have to surmount the entry barrier.The judge’s claim that there is “no” alterna-tive to Windows ignores, however, the obvi-ous fact that firms like Apple, Red Hat, Sun,and IBM (and several others) already haveoperating systems that can be used on per-sonal computers. They might not be “com-mercially viable” in the sense that they havesizable market shares, but that could bebecause Microsoft is not acting like a monop-olist. Firms like Apple, Sun, and Red Hat,which are already in the operating–systemmarket, could quickly become “commerciallyviable” (or more commercially viable thanthey already are) if Microsoft did whatmonopolists are supposed to do: restrictsales in order to raise the company’s pricesand profits. Few consumers would stay witha monopoly producer that could be expectedto charge monopoly prices far into thefuture, imposing on consumers costs thatthey could avoid by going with one of thealternative producers already in the market.

Granted, the judge also found that thereare “switching costs” for people who want tomove from one operating system to another.Sometimes, consumers of a new operatingsystem have to buy not only new softwarebut also new computers (as would berequired in any switch from aWindows/Intel-based personal computer toan Apple Mac). The consumers might alsohave to undertake training with the newsoftware and hardware. But that does notmean consumers will refrain from switch-ing. If a firm like Microsoft were to chargemonopoly prices, there would be “stayingcosts” associated with paying monopolyprices far into the future. The “staying costs”would equal the present value of payinghigher monopoly prices into the future.Hence, even if there are switching costs,those switching costs decline as the monop-olist raises its price. Computer users can beexpected to switch operating systems whenthe anticipated staying costs associated withan operating system are greater than theswitching costs.3

2

Judge Jackson hasfashioned a totally

new entry barrierthat has never

before been usedin antitrust cases,

dubbed the“applications

barrier to entry.”

Market Definition and the“Applications Barrier to

Entry”How then was the judge able to declare

that Microsoft is a monopolist in the operat-ing-system market? Very easily. He declaredMicrosoft to be in a narrowly conceived mar-ket, covering only single-user personal com-puters that come equipped with Intel orIntel-compatible microprocessors. That def-inition excludes many sales of operating sys-tems developed by Apple, Sun, Red Hat, anda number of other firms—either becausetheir computers use non-Intel-compatibleprocessors or because they are networked(i.e., linked to other users with whom theyshare resources).4 Given his narrowly definedmarket, Judge Jackson then found thatMicrosoft’s operating systems—includingDOS and all the dated versions of Windows,not just Windows 98—are installed on morethan 90 percent of the personal computersin the world, implying that Microsoft has“monopoly power,” or the ability to restrictsales and raise prices.5 Because there is “no”commercially viable alternative to Windows,according to the judge, no other producercould expand production to make up forMicrosoft’s sales reduction, which meansthat the price hike could stick.

Moreover, Judge Jackson found thatMicrosoft’s substantial market dominance, ashe defined the “market,” is protected by whathe chose to call an “applications barrier toentry.”6 He reasoned that the demand for anoperating system is dependent on the numberof applications written for that system. Thegreater the number of applications written foran operating system, the greater the value ofthe operating system to consumers. He alsofound that there are 70,000 applications writ-ten for Windows, a vastly greater count thanfor any other operating system.

As we shall see, Judge Jackson craftilyrigged the outcome of the trial in favor of thegovernment by making Microsoft a far moredominant force in the “personal computer”

industry than the company really is. He hasalso devised the required entry barrier behindwhich Microsoft, supposedly, exploited itsmarket power.

The judge asserts that Microsoft need notworry about normal competitive marketforces because existing or potential operat-ing-system rivals “would need a large and var-ied enough base of compatible applicationsto reassure consumers that their interests invariety, choice, and currency would be met tomore-or-less the same extent as if they choseWindows.”7 That is really a sly way of turningconsumer gains from using Windows againstMicrosoft. Microsoft has obviously done anoutstanding job of encouraging and helpingprogram developers to write for the “Wintel”platform, providing consumers with the“variety, choice, and currency” that theywant. But the judge argues that Microsoft’saccomplishment is the source of the monop-oly power that Microsoft can turn and hasturned against consumers by its supposedillegal and anti–competitive practices.

Several thousand software applicationswould not be sufficient to enable a potentialrival to erode Microsoft’s dominance, thejudge argues, because buying the alternativeoperating system “would still look like agamble from the consumer’s perspectivenext to Windows,” which, again, has 70,000applications.8 The proof of that statement,according to the judge, is evident in Apple’smarket predicament. Apple has 12,000applications written for its operating system,but “even an inventory of that magnitude isnot sufficient to enable Apple to present asignificant percentage of users with a viablesubstitute for Windows.”9 Hence, the judgeconcludes that Microsoft’s market domi-nance is protected because “the amount itwould cost an operating system vendor tocreate that many applications [70,000] isprohibitively large.”10 That entry barrierwould “prevent an aspiring entrant” fromdrawing away Microsoft’s customers even ifMicrosoft “priced its products substantiallyabove competitive levels for a significantperiod of time.”1 1

3

The judge con-cludes thatMicrosoft’s mar-ket dominance isprotected because“the amount itwould cost anoperating systemvendor to create70,000 applica-tions is prohibi-tively large.”

Where Judge JacksonWent Wrong

Judge’s Jackson’s first major mistake wasto reason that Microsoft’s market dominancegives it monopoly power worthy of antitrustscrutiny. In focusing on market share, as hasbeen traditional in antitrust cases over thepast century, the judge doesn’t seem to realizethat market share doesn’t mean as much in“new-economy” markets like software as itdoes in “old-economy” markets like steel. Inold-economy markets, the additional—or,more accurately, marginal—cost of expandingoutput is usually substantially above zero,because the production of more of a good likesteel requires firms to incur labor and materi-als costs, if nothing else.

In addition, the marginal cost of produc-tion can be expected to rise (within the rele-vant range of production and in the shortrun). That means that, if the dominant pro-ducer restricts sales to raise its prices andprofits, other firms in the market cannoteasily expand production in the near termwithout incurring significant additionalcosts. The ability of other firms to expandoutput to make up for the restricted sales ofthe dominant producer is ultimatelychecked by the rising marginal cost. At somepoint as production expands, the cost ofproducing one more unit rises above thegains the rival firms can get from selling theadditional unit, making further expansionno longer profitable. Hence, the dominantproducer’s restricted sales lead to a reduc-tion in total market supply and a higher(monopoly) price.

New-economy markets are quite different.As the judge found, software production mayrequire heavy fixed or up-front developmentcosts, but the marginal costs of producingadditional copies of a software product canbe “very small,” if not zero.12 A product likeWindows requires little in the way of addi-tional resources for, say, Dell Computers toelectronically copy Windows from the harddisk of its mainframe computer to the harddrive of each personal computer it builds.

Furthermore, the marginal cost of produc-tion can remain more or less constant at zero,or close to it, for as many copies as Dellchooses to produce.

Thus, mere dominance may mean little inthe operating-system market. If Microsoftdecides to restrict sales, some other existingoperating-system vendor (Red Hat or IBM),whose up-front development costs are also“sunk,” can make up for Microsoft’s restrict-ed sales at little or no additional cost, and thevendor can expand sales without confrontinga rising cost constraint. Even though he mis-takenly applied old-economy production eco-nomics to Microsoft and assumed thatMicrosoft’s market dominance gave it marketpower, Judge Jackson realized that he also hadto find a barrier to entry to make that marketpower durable.

Microsoft is clearly not as dominant in theoperating-systems market as the judge finds.The judge’s conclusion about Microsoft’sprotected market position can be disputed bythe simple observation that Apple may not bea dominant force in personal computers likeMicrosoft, but it certainly has given “a signif-icant percentage of users” not only “a viablealternative” to Windows but also a substitutemost Apple users would swear by. Yet Apple’ssmall market share in no way proves, or evenaffords credible evidence, that a potential rivalwould actually need more than 12,000 appli-cations to compete with Windows. Apple’smarket-share problems might be a conse-quence, not of Apple’s applications or lackthereof, but of some other more importantfactor—for example, Apple’s prices (whichhave always been higher than the prices ofcomparable Wintel personal computers),product design, or marketing strategy.1 3

It’s important to note that, in his findingsof fact, the judge gives no indication that heknows anything specific about the supposed“applications barrier to entry” other than thecount of applications, and surely the 70,000Windows applications (as well as the 12,000Apple applications) represent only a roughcount, which could be off by several thou-sand. No one seems to know exactly how

4

Apple’s smallmarket share in

no way proves, oreven affords cred-ible evidence, that

a potential rivalwould actually

need more than12,000 applica-

tions to competewith Windows.

many applications there are overall or howmany there are in various categories—forexample, the number of business (or produc-tivity) applications, games, or referenceworks. Many of the 70,000 applications maybe nothing more than utility programs thatare not likely to be sold as separate shrink-wrapped software packages. Indeed, it may bethat the count includes various editions ofgames, most of which are out of date and nolonger sold.1 4 The point is that the judgegives no indication that he really knows whatis in the barrier to entry on which he stakeshis theory of Microsoft’s monopoly power.

The 70,000 count of applications thatsupposedly form the barrier to entry appearsto come from the trial testimony of JohnRose, then head of Compaq Computers anda Microsoft witness. During the trial, he waspressed by Justice Department attorneyDavid Boies to explain why Compaq usedWindows. Rose noted that “approximately”70,000 applications had been written forWindows, and he agreed with Boies that theexistence of those applications was a “primereason” (Boies’ words) that Compaq shippedits computers with Windows on board. Lostin the judge’s “findings of fact” based onRose’s testimony is that some of the 70,000applications, according to Rose, were devel-oped for the DOS environment. Moreimportant, he stressed that there had been70,000 DOS/Windows applications devel-oped during the entire 17-year life of the per-sonal computer industry—not that there areactually 70,000 fully functioning, up-to-dateWindows applications currently available.1 5

Rose’s assessment is far removed from whatthe judge suggests when he asserts thatWindows “supports over 70,000 applica-tions” and concludes, therefore, that “theamount it would cost an operating systemvendor to create that many applications isprohibitively large.”1 6

Thus, the supposed applications barrierto entry is grounded in nothing more thanspeculation on the part of the judge concern-ing the role of the number of applications indetermining market share. His position is

certainly not founded on any courtroom evi-dence about the determinants of market sharefor operating systems. As a consequence, theinvestment a potential rival might have tomake in the development of applications isnot necessarily “prohibitively large,” eventhough the judge stakes his case for a breakupof Microsoft solidly on that claim. Surely, nopotential entrant into the operating-systemmarket would need or want to develop (orencourage the development of) counterpartsto DOS versions of applications, or evenWindows 3.0 versions. Applications writtenfor obsolescent operating-system technologiesare not likely to have much of a market.

In fact, the investment a challenger wouldhave to make in applications in order to takeaway at least a portion of the Windows mar-ket might actually be relatively modest, giventhat few computer users would ever beinclined to use hundreds, much less tens ofthousands, of applications even over a longstretch of time. Most people simply don’thave the time, need, or inclination to usemore than a handful of applications.17

The number of applications that all com-puter consumers together might use is also asmall fraction of 70,000. That fact is evidentfrom a search of the available productivityapplications sold on Amazon.com, which arelisted by several subcategories in Table 1.Amazon lists only 8,301 “software programs”under its productivity software categories.1 8

That is a large number, but it includes pro-grams written for all available operating sys-tems, including Red Hat Linux, IBM OS/2,and Apple Mac, as well as MicrosoftWindows. Many of the software programsare also listed several times in the varioussubcategories. Still, the unadjusted count ofproductivity programs represents just 12 per-cent of Judge’s Jackson’s 70,000 applications.

Moreover, the software market is besetwith “versioning,” the strategy of developing abasic product and then adding or subtractingfeatures, often at little or no added develop-ment expense, to generate several versions ofthe same product. Those versions aredesigned to fit the needs and wants of differ-

5

The judge givesno indicationthat he reallyknows what is inthe barrier toentry on which hestakes his theoryof Microsoft’smonopoly power.

Table 1Productivity Applications Listed on Amazon.com

Total TotalApplications Uniqueon Amazon Applications

Business and Office ApplicationsAccounting 95 34Career and management 97 58Communication utilities 1,346 171Database 93 32Document management 219 45

Office suites 95 14a

Personal finance 37 12Presentation 138 43Project manager 56 23Reports and forms 156 30Schedule and contact management 109 41Spreadsheets 25 2Training and tutorials 52 36Word processing 35 14Subtotal 2,553 555

Internet and Networking Applications 2,595 348

Graphics and Multimedia Applications3-D 73 32Animation 67 7CAD 103 31CD burning 15 9Clip art 148 73Image capture 2 2Desktop publishing 339 108Multimedia editing 84 54Illustrations 123 34Professional design 103 12Photo editing 155 51Video editing 71 18Subtotal 1,283 43

Utilities ApplicationsCross platform 28 5File compression 25 7File conversion 17 3Hardware and memory management 63 23Other 91 14PC maintenance 100 33Partition 14 2Screen savers 14 11Voice recognition 83 11Backup 517 88Virus protection 276 21Subtotal 1,228 218

Programming Applications 642 125

Total Productivity Applications 8,301 1,677

Note: Counts are based on listings between June 30 and July 6, 2000.

6

ent consumers in distinguishable markets.With versioning in software markets (as inother markets), application developers canprice discriminate by segmenting their mar-kets according to the buyers’ willingness to payfor different features, which can meangreater sales and profits.19

For example, a developer of office soft-ware can develop an array of productivityapplications, including a word processor,spreadsheet, database, Internet browser, slidepresentation, calendar, and Web page pro-gram. Then the developer can create moresophisticated or less sophisticated versions ofthose basic applications.20 In addition, thedeveloper can organize the basic applicationsinto an array of “office suites”—identified,possibly, as “standard,” “professional,” and“premier” or “deluxe”—with each versioncontaining different combinations of thefirm’s basic office applications, listed sepa-rately on Amazon and elsewhere.

To confuse the “count” of total applica-tions still further, software firms can thenprovide “add-on” programs, which offeradditional features (as in Windows Plus).Firms also can sell separate products for dif-ferent numbers of users (for example, 10, 25,50, or an unlimited number of users); each ofthose products may be counted, as they are inthe Amazon listing. The different versionsmay add to the firm’s packaging and market-ing cost but not necessarily to the cost ofsoftware development.

No one seems to know how John Rose gothis 70,000 count of software applications. Hesubmitted no documentation for the figurein his court testimony. Neither was Rose chal-lenged by either side to say where he got thenumber.2 1 Interestingly, the presumptionthat there may be tens of thousands, if not70,000 or even 100,000, Windows applica-tions may have originated with Microsoft,which may have been overzealous in its publicrelations and marketing efforts to accentuatethe value of Windows. Back in 1997, when itwas heavily promoting the adoption ofWindows NT for servers, Microsoft noted in apress release: “[H]undreds of leading applica-

tion vendors are writing rich and compelling32-bit Windows NT Workstation–readyapplications—with more than 100,000 32-bitWindows–based applications availabletoday.”22 Apparently, Microsoft has no docu-mentation to support the claim that there are“100,000 32-bit Windows-based applicationsavailable today.”

Without supporting documentation,it appears that the 70,000 applicationsthe judge elevates to the status of an“applications barrier to entry” includevarious versions of programs—for exam-ple, the standard and deluxe editions ofCorel’s WordPerfect Office Suite. Each versionof Office Suite includes, of course, underlyingproductivity applications that can also be soldand counted separately. Through versioning,basic productivity applications can be double-and triple–counted, and then some.Moreover, the Amazon count of 8,301 pro-grams includes upgrades of programs, aswell as full programs, and even tutorials. Thosemultiple counts and overlapping coverage,along with programs not normally thought ofas “applications,” greatly exaggerate the costthat operating-system and application devel-opers would have to incur in order to entertheir respective software markets.

To understand the extent of the duplica-tion involved in the Amazon count of businessand office programs, consider that Amazonhas 25 items for sale under its “spreadsheet”subcategory (which is itself under the “busi-ness and office” category), but there are onlytwo distinct spreadsheet programs, Excel andLotus 123, in the listing. There are 156 itemslisted under “reports and forms,” but againmany of the items are variations of the samebasic 30 programs. There are 95 “office suites”listed on Amazon, but only 13 suites whenproperly counted.

When only unique programs are counted,Amazon sells just 555 Windows-based appli-cations listed under the “business andoffice” subcategory in Table 1. When allother business subcategories (“internet andnetworking,” “graphics and multimedia,”“utilities,” and “programming”) are added,

7

When only uniqueprograms arecounted, Amazonsells just 555Windows-basedapplications in the“business andoffice” subcategory.

there are 1,677 unique Windows-based pro-ductivity applications listed on Amazon(many of which were, at the time of the survey,classified as “not in stock” or “not currentlyavailable”). That much more realistic count ofavailable Windows applications representsjust 2 percent of the judge’s “applications bar-rier to entry.”

Even then, that count exaggerates the bar-rier to entry in the desktop market. Morethan 400 of the included applications aredesigned exclusively for the back-office servermarket, which Microsoft does not dominateand was not an issue in the antitrust trial.Another 73 of the included applications areclip art programs that, with relative ease,could be adjusted to work with any new oper-ating system. Furthermore, it is not at allclear why a new operating system actuallyneeds 108 different desktop publishing pro-grams or 88 backup programs in order tomount an effective challenge to Windows. Allthat’s needed is a few very good options.

In short, there may be a thousand or soWindows-based productivity applicationslisted on Amazon that can reasonably count-ed as a part of the “barrier to entry”—no morethan 2 percent of the judge’s count.2 3

That means that there would have to be atleast 68,000 nonproductivity applications forthe total count to match the judge’s estimateof the “applications barrier to entry.” Yetthere are only 3,644 combined applicationslisted on Amazon under the separate cate-gories of “children’s programs” (1,458), “edu-cation and reference” (416), “games” (1,480),and “homes and hobbies” (290). That totalcount exaggerates the number of uniqueapplications because many applications arelisted several times under the subcategories,and versions and upgrades of gamesabound.24 Also, an undetermined number ofthe programs in those categories (especiallyin “education and reference”) do not have tobe fully recoded for a new operating system,which means that a promising new entrantinto the operating-system market couldexpect to add those programs with relativeease.

Amazon, of course, does not sell allWindows-based applications. Also, manyWindows applications are specific to indus-tries and even firms. At the same time,Amazon must sell a sizable percentage of allavailable Windows applications, especiallythe widely used programs, and the Amazonlist includes many programs that only com-puter technicians and experts might buy (forexample, Tabworks, Faxrouter, Hostexplorer,and Cosession Remote). No doubt, the arrayof Windows programs makes entry less thaneasy, as should be expected. Building theWindows application base was not easy tobegin with. Microsoft has invested billions inbuilding its applications network by helpingdevelopers write for Windows.

The point is that Judge Jackson has stakedhis case against Microsoft on the explicit andvery precise claim that there are 70,000Windows-based applications, which, becauseit is a large number, necessarily forms a “pro-hibitive” barrier to entry. In staking out hisposition so firmly and precisely, he has gross-ly exaggerated the entry barrier, a point thathe would surely have conceded had he knownmore about what was in the applicationscount that he accepted as legal “fact.” Becausethe array of Windows applications is farsmaller than what he claimed, the requiredinvestment for entry must be far less than heindicated. Granted, the investment requiredto challenge Microsoft could still be substan-tial (given the complexity of a number of theapplications in the final count), but that is afar cry from “prohibitive.” Besides, is it notreasonable to expect new firms to incur manyof the same development costs that Microsofthad to incur?

It looks as though the up-front invest-ment would be well within the means of ahost of existing and potential software firms,if Microsoft were to act like a monopolist.That is especially true since operating-systemchallengers have to fund only the effort toencourage program development by others.That is, they would not have to fund theactual writing of the applications themselves.The applications firms would do that. Even

8

It is not at all clearwhy a new operat-

ing system actuallyneeds 108 different

desktop publish-ing programs or

88 backup pro-grams in order to

mount an effectivechallenge to

Windows.

then, given the prevalence of “networkeffects”—which were at the foundation of thegovernment’s case against Microsoft andwhich Judge Jackson adopted as “fact”—an“aspiring entrant” would have to spur thedevelopment of only a few applications at thestart.2 5 Those initial applications can beexpected to cause an increase in sales of thechallenger’s operating system, which can giverise to more applications written by outsidedevelopers, which can lead to even moreoperating-system sales and a greater numberof applications. Again, encouraging thedevelopment of the “network” would be espe-cially easy for an “aspiring entrant,” ifMicrosoft were to act like a monopolist.

Applications Used byMBA Students

The fact that most computer users typi-cally use few applications is obvious from asurvey of 152 working managers enrolled inthe part-time MBA programs at theUniversity of California, Irvine.2 6 All surveyparticipants were computer literate and usedcomputers on their jobs and in their graduatestudies. If there is a group of people whoshould be expected to use a wide variety ofprograms, it is that one. Still, when askedhow many different computer programs(games included) they had used over the fivemonths before the survey, 19 percent of therespondents noted that they had not usedprograms other than the six that come withthe Microsoft Office suite (which came pre-loaded on their university-issued laptops thatwere required for their classes). As indicatedin Table 2, 42 percent used only one to threeprograms in addition to Office. Nearly 22percent used four to six additional programs,and only 16 percent used more than six addi-tional programs. That means 84 percent ofthe respondents used in total a dozen orfewer applications (including Office applica-tions and games). No one reported usingmore than 18 programs—Office plus 12other applications.

Of course, there was significant variation inthe programs used. Respondents reportedusing 187 different productivity applicationsat work (ranging from 3-D CAD to BaconsMedia to Lotus Notes, as well as otherWindows-based programs developed exclusive-ly for their firms). They also reported using 47different games or entertainment programs(ranging from Delta Force to Sim City toUnreal Tournament). The survey respondentsmay not have accurately remembered all of theprograms they had used. The count of pro-grams used would very likely rise with a length-ening of the time frame covered by the survey,meaning the findings may understate the vari-ety of programs a firm would need to domi-nate the operating-system market.

Clearly, Microsoft has a strong marketadvantage because of the many applicationswritten for its operating system. At the sametime, it is obvious that a sizable segment ofcomputer users—maybe a fifth (or more,given that the survey was distributed to userswho are likely to be far more computer savvythan the general population of users)—couldbe adequately served with an operating sys-tem that has a few dozen applications writ-ten for it. Surely, a few hundred well-chosenapplications might be all that would beneeded for a new entrant to surmount the“applications barrier” and to make a crediblechallenge to Windows.

A fifth of the U.S. market covers somethingon the order of 20–25 million computer users.A fifth of Microsoft’s operating system rev-enue amounts to about $2 billion annually.That seems like a fairly sizable market, whichany number of firms would be eager to enter ifMicrosoft were ever to act like a monopolist—that is,charge monopoly prices that would make the cost ofstaying with Windows prohibitive. Then,Microsoft would surely be vulnerable to entry.

Sun Microsystems obviously doesn’tbelieve that tens of thousands of applicationsare needed to make headway in the operat-ing-system market. Sun is seeking to becomea network “application services provider” (asare Microsoft and several other firms). Sunintends to rent computing services over the

9

A sizable segmentof computerusers could beadequately servedwith an operatingsystem that has afew dozen appli-cations writtenfor it.

Web and charge by the use. Toward that end,Sun started its ASP venture with the eightbusiness applications acquired in its 1999buyout of Star Division. To build its businessbase, it has been doing what Microsoft didwith Internet Explorer: giving away its soft-ware products.

The Absence ofMonopoly Pricing

A second problem with the judge’s find-ings is that nowhere in his analysis of theoperating-system market does he find thatMicrosoft has actually raised its price “sub-stantially above competitive levels for a sig-nificant amount of time.” Microsoft’s pricefor Windows is at or below the prices com-petitors like IBM, Sun, and Apple charge fortheir operating systems. Moreover, the priceof Windows has fallen in real (inflation-adjusted) terms by over 50 percent over the

past decade, without adjusting for productenhancements.27 Now, Windows 98 upgradessell at retail for $89, but that figure overstatesthe cost of Windows to nearly 90 percent ofPC users, who get their operating systemwhen they purchase a new computer.Microsoft sells copies of Windows to com-puter manufacturers for approximately $65(or less). The Graduate School ofManagement at the University of California,Irvine, is able to license copies of Windows98, plus all of the applications in Office 2000and Front Page (several hundred dollars’worth of software when bought at retail), for$47.50 a year for faculty and staff and $18.60a year for MBA students. Do those sound likethe prices that a monopolist, protected by a“prohibitive” barrier to entry, would beexpected to charge?

Indeed, Judge Jackson never found evi-dence of monopoly pricing. Instead, he con-cluded: “The debut of Internet Explorer andits rapid improvement gave Netscape an

10

Judge Jacksonnever found evi-

dence of monop-oly pricing.

Table 2Distribution of Windows Applications Used by Fully Employed (and part-time)MBA Students

Percentage ofApplication Students

Only the applications in Microsoft Office 19%

1–3 applications in addition to Microsoft Office 42%

4–6 applications in addition to Microsoft Office 22%

More than 6 applications in addition to Microsoft Office 16%

Number of different work-related applications used by all respondents 186

Number of different entertainment or education applications used by all respondents 47

Note: Survey undertaken by the author during June 2000 at the Graduate School of Management,University of California, Irvine.

incentive to improve Navigator’s quality at acompetitive rate. The inclusion of InternetExplorer with Windows at no separate chargeincreased general familiarity with the Internetand reduced the cost to the public of gainingaccess to it, at least in part because it com-pelled Netscape to stop charging forNavigator. Those actions thus contributed toimproving the quality of Web browsing soft-ware, lowering its cost, and increasing itsavailability, thereby benefitting consumers.”28

Why did Microsoft do all of those goodthings? According to the judge, “to protectthe applications barrier to entry, and hence itsmonopoly power.”2 9 But the issue the judgedoesn’t address is why Microsoft would wantto protect the kind of “monopoly power” thatrequires it to upgrade Windows to includebrowsing capability and then to give thatadded benefit away. The usual barrier to entrywould allow the protected firm to restrictsales with the intent of raising its prices. The“applications barrier to entry” doesn’t soundlike much of a “monopoly barrier.” Still, thejudge managed to find that Microsoft’s tac-tics caused “consumer harm by distortingcompetition.” Is not responding to the entryof a rival with improved products and lowerprices the very essence of competition?

Nevertheless, Judge Jackson concludedthat the “predatory course of conductMicrosoft has pursued since June of 1995 [inthe main, integrating Internet Explorer intoWindows and then giving away the browserfunctionality] has revived the dangerous proba-bility that Microsoft will attain monopolypower” in the browsing market.3 0 But sowhat if Microsoft attains “monopoly power”if it uses that supposed power to improve the“quality of Web browsing software, loweringits cost, and increasing its availability, therebybenefitting consumers”?31 In addition, thejudge fails to acknowledge that Microsofthas helped to break down whatever applica-tions barrier to entry there is by rewriting itsown Office applications for the Apple Macoperating system, thus ensuring that its ownapplications will not stand in the way of Mac’staking over a larger share of the market.32

In essence, when the judge’s findings offact and conclusions of law are stripped oflegal parlance, it is clear that he really foundthat Microsoft charged too little, not toomuch, for Windows. Remember, he foundthat the zero price for the integrated browser,Internet Explorer, was “predatory,” designedto ruin Netscape as a viable competitor.3 3

Those pricing facts are hard to reconcile withthe judge’s claim about Microsoft’s monopolypower because any self-respecting monopolistwould take advantage of its market power byrestricting sales and raising prices. If the cur-rent owners have no interest in doing so, thensurely other savvy investors would be willingto buy a controlling interest in Microsoft andchange its pricing policies—if Microsoft had themarket protection the judge thinks it has.

The charge that Microsoft could act like amonopolist in the future has been leveled forat least a decade, yet all the while the realprice of Windows has continued to fall. Onemight think the judge would wait to con-demn Microsoft until it had actually raisedits prices above competitive levels beforeproposing to break up the company andrestrict its ability to give away the browser.One might also think the judge would under-stand this basic proposition: If Microsoft hadto respond with such force and a zero price toNetscape’s market entry—a Web browser thatwas written in four months by seven recentgraduates of the University of Illinois—theapplications barrier to entry must not bewhat he made it out to be.

Worries about the NearTerm and the Long Run

At several points in his analysis ofMicrosoft’s behavior, the judge is clearly wor-ried about what Microsoft might do in thenear term, over the next two or maybe threeyears.34 He obviously believes that, with itssupposed protection from normal competi-tive forces, Microsoft has the requisite marketpower to raise its prices substantially by cur-tailing sales. Microsoft might in fact have such

11

If Microsoft hadto respond withsuch force and azero price toNetscape’s mar-ket entry, theapplications bar-rier to entry mustnot be what thejudge made it outto be.

power—in the near term. After all, many com-puter users would be hard-pressed to switch toanother operating system in the near term,but that in no way means that Microsoft hasthe requisite incentive to raise its prices in thenear term. That is because near-term pricehikes can cut sales. Moreover, as Judge Jacksondetermined, software markets are beset with“network effects,” or the tendency of sales tobuild on sales with time.35 If more copies ofWindows are sold, application developers havemore reason to develop programs forWindows. With more applications, the valueof Windows rises along with sales.

Network effects can also work in reverse.A reduction in near-term sales due to a priceincrease can leave a market opening for someother operating system that can establish abeachhead with lower prices leading togreater sales, which attract application devel-opers. More applications can lead, in turn, tomore sales of the new operating system atMicrosoft’s expense. Therefore, Microsoftmight be able to raise its profits in the nearterm, but those near-term profits wouldentail an even greater loss of profits in thelong term. As a consequence, a price increasefor Windows today could translate into areduction in the price of Microsoft’s stock.The current stock price should reflect thestream of expected future earnings, whichcould well be lower than they would be ifMicrosoft did not raise its current price forWindows. One reason Microsoft has notbehaved as if it is protected by the “applica-tions barrier to entry” is that the barrier issimply not what the government and judgepretend it is. Microsoft’s key employees alsohave their sights focused on the company’sstock price (because of their stock holdingsand stock options), not on its current bot-tom line. How else can a retail price of $89for Windows 98 or an annual lease paymentfor students of $18.60 for Windows andOffice be explained if Microsoft is such adominant producer camped out behind a“wall” of applications?

Seen from that perspective, the “barrier toentry” that is protecting Microsoft is not the

applications but the prices it charges.Because Microsoft constrained the price ofits operating system (say, relative to the pricecharged by Apple), program developers werewilling to write for Windows, which led tothe much-heralded “network effects” (i.e.,sales of Windows resulting in more applica-tions that, in turn, led to more copies ofWindows being sold).

In effect, Judge Jackson has, with thestroke of his pen, converted the 70,000 appli-cations, which consumers value and use, intoa source of consumer harm caused byMicrosoft’s supposed anti-competitive tac-tics. And what did Microsoft do? Well, it did-n’t prevent anyone from using any of thosesupposed 70,000 applications. There hasnever been a moment when consumers couldnot use Netscape Navigator with Windows.All Microsoft has ever done is upgradeWindows by integrating its browser, InternetExplorer, into Windows. Then it gave awaythe browser functionality by not raising theprice of Windows. In the process, Microsoftforced Netscape to lower the list price forNavigator from $39 before Microsoft enteredthe browser market to zero afterwards.

By integrating a browser into Windows,Microsoft gave practically every user of Intel-compatible and Apple computers the tech-nology necessary to go to Netscape’s Web siteand download Navigator for free. In doing so,Microsoft may have actually stimulated theuse of Navigator, especially since Microsoftmade browsing virtually ubiquitous and,hence, encouraged the speedy development ofthe Internet world, as the judge acknowl-edges. In short, Microsoft gave more peoplemore reason and more ability to downloadand use Navigator.

Granted, after Microsoft entered the so-called browser war, Netscape lost marketshare, as practically everyone knows. Butwhat is not so widely known is that the actu-al count of Navigator users exploded duringthe two years of the trial from 15 million in1996 to more than 35 million at the end of1998, the period during which Microsoft wassupposedly foreclosing Netscape from the

12

Judge Jackson has,with the stroke of

his pen, convertedthe 70,000 applica-

tions, which con-sumers value anduse, into a source

of consumerharm caused by

Microsoft’s sup-posed anti-com-

petitive tactics.

browser market. And the number ofNetscape users may have doubled again sincethe trial started in May 1998.36

In any event, Netscape’s NetCenter pro-vides a variety of e-commerce services and hasbecome a major hub of Web-based economy,which partially explains why America Onlinewas willing to pay more than $10 billion forNetscape while the Microsoft antitrust trialwas going on. A part of that purchase pricemight have been attributable to Microsoft’shelping to make browsing ubiquitous in away that Netscape alone could not have donein the same time frame. And Windowsallegedly dominated the operating-systemmarket because of those supposed 70,000applications. That means the so-called appli-cations barrier to entry might have been seen,contrary to the judge’s imagery, as a “con-duit” through which Netscape was able toflourish because the Web developed morerapidly than it would have otherwise.

We can’t be certain, nor can the judge,whether Microsoft’s supposed “predatory”actions were, on balance, a boon or a bane toNetscape. No evidence was introduced at trialthat spoke to that empirical issue. But we doknow one thing for sure: Microsoft did notpredatorily (or anti-competitively) driveNetscape from the market.

Consumer Gain from the“Applications Barrier to

Entry”Judge Jackson doesn’t seem to realize that

the mere existence of 70,000 Windows appli-cations (as if there really were that many)indicates that Microsoft must have eased theentry of hordes of program developers intothe market for Windows-based applications.In doing so, the company, no doubt, stirredcompetition among a multitude of applica-tion firms, leading to product improvementsand falling prices, and thus raising the valueof Windows and personal computers tousers. And, by elevating the value of theWintel platform, Microsoft spurred competi-

tion among computer makers, ensuring thatthe price of computing would also decline.

Even if Microsoft has been chargingmonopoly prices (which, again, has not beendemonstrated),37 it does not follow thatMicrosoft has, on balance, harmed consumers.The reduction in the prices of Windows appli-cations resulting from the ease of entry intothat market, as well as the decrease in the priceof computing, may have more than offset anyincrease in the price Microsoft might havecharged for Windows itself. Had Microsoftnot been able to gain in some way from itsencouragement of the development of appli-cations that supposedly form a “barrier” tocompetition, there might not have been asmuch competition in applications and com-puters as there has been. The market mighthave been dominated by the Apple model ofcomputer markets, under which the operatingsystem and computer are sold as a tied pack-age—and at higher prices than comparablepersonal computers running Windows.

The judge speaks of the applications bar-rier to entry as if it came into existence moreor less naturally, as a consequence of net-work effects that are, like gravity, not theproduct of human ingenuity and effort.That is hardly the case. Getting independentdevelopers to write for any operating systemis not a consequence of natural forces, nor isit a costless undertaking. Indeed, Microsofthas spent billions of dollars facilitating thework of software developers. No companywould undertake such an investment if it didnot expect to get something in return. By thesame token, if programmers expected to betrapped by a monopolist that would chargemonopoly prices once they had written theirapplications, they certainly would be lessinclined to write programs. That means afirm like Microsoft would have to investmore resources (or charge lower prices) inorder to get the programmers’ cooperation.The threat of Microsoft’s charging monop-oly prices in the future could similarlyencourage developers to write for alternativeoperating systems, which could open up themarket to new entrants.

13

Microsoft stirredcompetitionamong a multi-tude of applicationfirms, leading toproduct improve-ments and fallingprices, and thusraising the value ofWindows and per-sonal computersto users.

Microsoft has a history of charging lowprices, then lowering its prices still further andenhancing Windows over time.3 8That helps tomake credible Microsoft’s commitment toremain the operating-system standard. Apple,by contrast, has far fewer applications eventhough Apple’s graphics-based operating systemis much older. It is likely that Apple has investedfewer resources in encouraging program devel-opers. Moreover, Apple has a reputation forbeing pricey. Perhaps those reasons—not theapplications barrier to entry—account forApple’s role as a minor player in the personalcomputing market.

Judge Jackson appears convinced thatgreater consumer mobility among operatingsystems would be a boon to competition andconsumer welfare. But is that necessarily thecase? If consumers could move among operat-ing systems with complete ease, it is doubtfulthat operating-system companies would bewilling to work as hard at getting programdevelopers to write applications. Programmersmight not be so willing to make the invest-ments they have to make. Hence, the so-calledapplications barrier to entry in the operating-system and applications market can be seen asa boon to the development of applications. Anapplications barrier of some magnitude cangive operating-system companies and programdevelopers the incentive they need to entertheir respective markets.3 9

Moreover, program developers are not asunsophisticated as the judge implies when hewrites that they are trapped behind the appli-cations barrier to entry. They understand thatthe value of their applications is related tohow widely Windows is used. Like Microsoft,they understand that the success of theWindows standard is related to its price. Ifthey expected Microsoft to charge monopolyprices in the future, they would be less likelyto write for Windows, which means that themere threat of Microsoft’s acting like amonopolist can undermine the supply ofWindows applications. The fact thatMicrosoft has so many applications writtenfor Windows suggests that the company hasbeen highly successful in convincing pro-

grammers that it will not charge prices so outof line with competitors’ prices that sales willbe inhibited. In other words, the manyWindows applications (whatever the count)are a mark of Microsoft’s competitiveness. IfMicrosoft did not show that it could andwould remain competitive, vis-à-vis alterna-tive operating systems, it would not have thelarge count of applications and, accordingly,would not have achieved its dominant posi-tion in the operating-systems market.

Because Microsoft integrated InternetExplorer into Windows and then did notcharge for the browser functionality, JudgeJackson concludes that Microsoft had “preda-tory” intentions, in the main, to destroyNetscape. According to the judge, “[P]redatoryconduct, by definition as well as by nature,lacks pro-competitive business motivation.”4 0

Microsoft’s efforts with its browser may havebeen intended to harm Netscape, but suchharm should be expected in competitive mar-kets. Competitors are destroyed all the time.

Moreover, Microsoft’s “motivation” mayvery well have been pro-competitive. Microsoftmay have understood that the viability of theWindows standard, and all its applications,depended on ensuring that Windowsremained ubiquitous. It would not haveremained so if Windows did not have browsingcapability. The program developers could alsosee what Microsoft saw: If Microsoft had notincluded a browser in Windows, many pro-gram developers would have turned away fromWindows, thus undermining the acclaimednetwork effects and causing the Windows stan-dard to unravel. Microsoft’s zero pricing ofInternet Explorer is one way Microsoft alertedthe program developers that it intended toremain competitive in the face of a strong chal-lenge, a stance that fortified the Windows stan-dard (which the judge mistakenly interpretedas anti–competitive behavior).

In effect, Microsoft could be said to haveused the aggressive tactics it did to defend theinterests of developers—not because Microsoftis altruistic, but because developers’ interestsoverlap Microsoft’s own long-run interest. It ishard to see how a software firm can get more

14

Microsoft could besaid to have used

the aggressive tac-tics it did to defend

the interests ofdevelopers—not

because Microsoftis altruistic, but

because develop-ers’ interests over-

lap Microsoft’sown long-run

interest.

“pro-competitive” than that. Maybe that’s whyMicrosoft’s rivals have been so critical ofMicrosoft. Had Microsoft acted like a truemonopolist—one that restricted sales to raiseits prices—rivals would have applaudedbecause they, too, could then have made moresales at higher prices.

How Windows ApplicationsCan Open the Operating-

System MarketFinally, Judge Jackson (and just about

everyone else who agrees with his decision) hastreated the array of Windows applications as ifit were an unmitigated blessing for Microsoft.Without question, as recognized throughoutthis paper, those applications are an impor-tant marketing advantage for the company. Atthe same time, when considering ways toupgrade Windows, Microsoft must be evermindful that it cannot move too quickly in itsupgrades, or else it runs the risk of disaffectingits program developers who have investedheavily in Windows technology, a problemthat has important but easily overlooked neg-ative consequences for Microsoft.

Foremost, Microsoft must restrict (to onedegree or another) its upgrades of Windows,as it has done for years in not moving com-puter users to what the company knows to besuperior, more stable operating-system tech-nology, for example, Windows NT (now2000). If Microsoft readily makes existingapplications obsolete with each new releaseof Windows, it increases the computing costsfor consumers, given that consumers thenmust buy new versions of their applicationsor work without as many applications. Thecompany would also increase the costs todevelopers, since developers would then haveto rewrite their applications with every newWindows release. As a result, Microsoftwould have to invest more to get programdevelopers’ cooperation or would have tolower the price of Windows to keep their alle-giance to the Windows platform, or both.That implies that the economics of software

development can militate against Microsoft’smaking Windows as good as it could be.

Another way of stating the problem is thatMicrosoft must ensure that Windows is back-ward compatible, meaning new versions ofWindows can also run applications that werewritten for earlier versions of Windows (say,Windows 95, if not Windows 3.0). Therefore,Windows will always have many more lines ofcode than would be required for it to run onlythe up-to-date versions of applications.Moreover, the current version of Windows willbe more complex than would be necessary ifWindows did not have to be backward com-patible. Those problems explain why Windowsis less stable, perhaps slower, than it could be,absent the backward compatibility requirement.

Therein lies the potential for a marketopening for new entrants who are unencum-bered by having an operating system basedon old technology or who do not have toaccommodate an array of existing applica-tions. As a consequence, Microsoft has toweigh the ability of new entrants to leapfrogits established operating-system technology inits decisions to exploit what Judge Jacksonsees as market power. It must also stand readyto radically revise its operating-system tech-nology in relatively short order when it detectsa new technology that shows promise ofleapfrogging Windows (which may explainthe company’s abrupt change in its operating-system development plans when Bill Gatesopenly recognized the Internet browser threatin May 1995).41 Microsoft’s radical revisionscan be viewed as protecting not only its owninvestment but also the investments of con-sumers and application developers.42 IfMicrosoft were to act like the monopolist ithas been accused of being, it would give poten-tial new entrants with new technology all themore incentive to invade the operating-systemmarket.

Conclusion

Judge Jackson’s use of the word “barrier”conjures up the image of an impregnable

15

Had Microsoftacted like a truemonopolist—onethat restrictedsales to raise itsprices—rivalswould haveapplaudedbecause they, too,could then havemade more salesat higher prices.

fortress. Ironically, Microsoft stands con-victed of antitrust violations partiallybecause of the prominence of “applicationsbarrier to entry” in the trial. On close exam-ination, it is not the fortress the judge hasmade it out to be. If it were, Netscape couldnot have challenged Microsoft. Microsoftwould not have been compelled to respondto Netscape’s challenge in the way and tothe extent it did. Microsoft could haveremained content behind its impregnablebarrier. It might never have upgradedWindows to include browsing capability,and it would certainly not have given thatcapability away.

If Microsoft were the protected monop-oly it has been made out to be, it would havecharged far more than it has for Windows.There would then have been real harm toconsumers, who would have bought fewercopies of Windows at inflated prices. Therewould then have been real antitrust viola-tions. If Microsoft had acted like a protect-ed monopolist, it might never have achievedits dominant market position and wouldnot have been charged with antitrust viola-tions. Microsoft’s market rivals would haveapplauded its sales restrictions and above-competitive prices because those rivalswould have been able to sell more operatingsystems at higher prices.

The reason that Microsoft has chargedno more for Windows than it has, andresponded the way it did to the Netscapechallenge, is that the so-called barrier toentry made up of 70,000 (or 100,000) appli-cations is a gross exaggeration of just howmany up-to-date, available Windows-basedapplications are in use. Thus, Judge Jacksonerred in a major way when he concluded that“the amount it would cost an operating sys-tem vendor to create that many applications[70,000] is prohibitively large.” If Microsofthad acted like the monopolist the courtdeclared it to be, and if Microsoft’s users haddemanded the 70,000 applications thatJudge Jackson posited, then competitorsmight indeed have been barred. Fortunatelyfor consumers, none of that is true.

NotesThe author is indebted to Gary Byrne, KennethElzinga, Robert Levy, Albert Nichols, AlanReynolds, and William Shughart for their com-ments and editorial improvements on earlier ver-sions of this paper.

1. Judge Jackson seems to have taken the conceptof the “applications barrier to entry” from DavidBoies, the government’s lead attorney on the case,who noted in his opening statement before thecourt: “In fact, what is being in the driver’s seathere is the desire to shut out the NetscapeNavigator, and as the court will see when we talkabout AOL, they were prepared to do that, even atthe extent of disfavoring their own MSN networkin competition with AOL, because winning thebrowser battle was of paramount important andwas of paramount importance, I will repeat, fortwo reasons. One, it represented an ability toreerect the applications programming barrier toentry and, second, it allowed them to gain controlof the browser, which was, as they have themselvesrecognized, the choke-hold on the Internet.”United States v. Microsoft Corporation, U.S. DistrictCourt for the District of Columbia, civil actionno. 98-1232 (TPJ), Transcript of Proceedings, June7, 1999 (A.M. session), p. 61. Cited hereafter asTranscript of Proceedings. In turn, Boies appearsto have taken the concept from the direct testi-mony of the government’s two economic witness-es, Franklin M. Fisher and Frederick Warren-Boulton. Fisher called the concept the “applica-tions programming barrier to entry”: “For exam-ple, if Internet browsers and/or Java in fact threat-ened to eventually undermine Microsoft’s operat-ing system monopoly by eroding the applicationsprogramming barrier to entry that protects thatmonopoly, the economic costs of permittingMicrosoft to rebuild that barrier to entry by sti-fling non-Microsoft browsers and Java will besubstantial.” United States v. Microsoft Corporation,Franklin M. Fisher, Direct Testimony, filedOctober 14, 1998 http://www.usdoj.gov/atr/cases/f2000/2057.pdf.

Similarly, Warren-Boulton testified: “As anoperating system gains popularity, the incentiveto develop software for that operating systemincreases because the larger number of users forthe operating system product implies a greaterpotential market for software developers. Thedevelopment of yet more applications for thatoperating system, in turn, increases the value ofthe operating system to end users who, asexplained, purchase operating systems in signif-icant part based upon the quality and variety ofapplications available for it. As Hewlett-Packard’s Frank Santos explained, demand foran operating system is driven by the availability

16

If Microsoft’susers had

demanded the70,000 applica-

tions that JudgeJackson posited,

then competitorsmight indeed

have been barred.Fortunately for

consumers, noneof that is true.

of ‘applications that run on the operating sys-tem.’ The operating system’s market share,therefore, is likely to increase, and that, in turn,is likely to cause software developers to devoteyet more resources to writing applications forthat operating system product. That phenome-non—known in economics as ‘positive feed-back’—creates what is best termed the ‘applica-tions barrier to entry.’ Simply put, an operatingsystem product can rise to dominate the mar-ket, and once that dominance is achieved main-tain it, because of both the large number ofcomplementary software applications availablefor it and the flow of new applications that arewritten to it.” United States v. MicrosoftCorporation, Frederick R. Warren-Boulton,Direct Testimony, filed November 18, 1998, ¶ ¶53–54, http://www.usdoj.gov/atr/cases/f2000/2079.htm#footbody_27.

Interestingly, while Fisher and Warren-Boulton stake their claim of Microsoft’s monop-oly power on the existence of the “applicationsbarrier to entry,” neither ever offers any evidenceon exactly how many applications are in the bar-rier to entry. The closest either gets to an exactcount is when Warren-Boulton wrote thatMicrosoft was intent on “preserving the barrier toentry created by the large stock of Windows appli-cations” (¶ 7, emphasis added). He adds later,“[B]arriers to effective entry into the PC operatingsystem market are high, and other PC operatingsystems cannot easily increase their shares of thatmarket because the huge stock of applicationswritten for Windows 95/98 will not run on thosesystems” (¶ 43).

2. United States v. Microsoft Corporation, Findings ofFact, November 5, 1999, ¶ 54. Cited hereafter asFindings of Fact.

3. Computer users could incur the switching costsbut then effectively cover those costs with reduc-tions in the cost of their software.

4. For more details on Judge Jackson’s narrow mar-ket definition, See Alan Reynolds, “U.S. v.Microsoft: The Monopoly Myth,” Wall StreetJournal, April 19, 1999, p. A12.

5. Judge Jackson wrote: “Every year for the lastdecade, Microsoft’s share of the market for Intel-compatible PC operating systems has stood aboveninety percent. For the last couple of years the fig-ure has been at least ninety-five percent, and ana-lysts project that the share will climb even higherover the next few years. Even if Apple’s Mac OS wereincluded in the relevant market, Microsoft’s sharewould still stand well above eighty percent.”Findings of Fact, ¶ 35. That finding fortifies his ear-lier finding: “It follows that, if one firm controlledthe licensing of all Intel-compatible PC operating

systems world-wide, it could set the price of a licensesubstantially above that which would be charged ina competitive market and leave the price there for asignificant period of time without losing so manycustomers as to make the action unprofitable.Therefore, in determining the level of Microsoft’smarket power, the relevant market is the licensingof all Intel-compatible PC operating systems world-wide.” Ibid., ¶ 18. What the judege doesn’t say isthat Microsoft’s 95 percent “market share” for the“last couple of years” is not a fact at all. It was intro-duced early in the trial as a projection of Microsoft’sinstalled base for the next couple of years. The 95percent figure bears no relationship to current orfuture sales. There is no market for DOS operatingsystems, a point emphasized by Reynolds. In a per-sonal communication on July 21, 2000, Reynoldsinformed the author that the International DataCorporation puts Microsoft’s market share of cur-rent operating system sales, not installed base, atcloser to 66 percent.

6. Findings of Fact, ¶ 31.

7. Ibid., ¶ 40.

8. Ibid.

9. Ibid., ¶ 47.

10. Ibid., ¶ 40.

11. Ibid., ¶ 36.

12. Judge Jackson found: “Software development ischaracterized by substantial economies of scale.The fixed costs of producing software, includingapplications, are very high. By contrast, marginalcosts are very low. Moreover, the costs of develop-ing software are sunk—once expended to developsoftware, resources so devoted cannot be used foranother purpose. The result of economies of scaleand sunk costs is that application developers seekto sell as many copies of their applications as pos-sible. An application that is written for one PCoperating system will operate on another PC oper-ating system only if it is ported to that system, andporting applications is both time-consuming andexpensive.” Ibid., ¶ 38.

13. Indeed, Apple’s market share may now be“small” simply because it has tried to act more likea monopolist than Microsoft has. Its premiumprices for both hardware and software could havecurbed Apple’s sales. And those reductions in salescould have had “network effects,” checking thenumber of applications written for Apple and thussubsequent sales. In addition, the very fact thatApple has charged premium prices could havecaused both computer users and developers justifi-able concern about joining the Apple network.

17

Users and developers may have had reason tobelieve that Apple would, if it became the standard,act even more like a monopolist in the future, soak-ing up the gains the users and developers mighthope to realize by joining the Apple network.

14. Many of the Windows-based games countedas part of the applications barrier to entry arerapidly becoming obsolete because Internet-basedgames offer multiuser features not available onsingle-user computers.

15. At one point, Justice Department attorneyDavid Boies pressed John Rose, “Now, back to thequestion I started with, there are 70,000 applica-tions, approximately, for the Windows operatingsystem; correct?” Rose then explained, “There are70,000 PC applications. Not all of them were writ-ten for Windows. Some of them go back to theDOS environment and carry forward into theWindows environment. But over the course of the17 years of the PC industry, there have beenapproximately 70,000 applications developed,and that continues to grow.” Transcript ofProceedings, February 17, 1999 (P.M. session), p.24.

16. Findings of Fact, ¶ 40.

17. No computer user, no matter how dedicated,would ever attempt to use 70,000 applications(even if that many usable, unique applicationswere available, which is not the case) over thecourse of his or her life. If it took an average of 30minutes for a user to install and learn each appli-cation, it would require 35,000 hours to installand learn all 70,000 Windows applications. If theuser spent 40 hours a week working only on newapplications, it would take 109 years to install andlearn them all.

18. Productivity software includes all applicationsexcept children’s, educational and references,homes and hobbies, and personal digital assistantsoftware. See http://www.amazon.com/exec/obidos/tg/browse/-/289961/104-4623710-6770356.

19. For more on the economics of versioning, seeCarl Shapiro and Hal R. Varian, Information Rules:A Strategic Guide to the Network Economy(Cambridge, Mass.: Harvard Business SchoolPress, 1999), chap. 3.

20. The various versions of software might beequated with given car models with and withoutluxury or sports packages.

21. Rose’s testimony may well have gone unchal-lenged by Microsoft’s lawyers because he was theirwitness. To challenge him on the count of appli-

cations could have undermined his credibility onother points. By leaving the count unchallenged,however, Microsoft gave the government anopportunity to repeat the 70,000 count severaltimes, which gave the judge all he needed to findthe entry barrier that was necessary to make themonopoly charge stick.

Richard L. Schmalensee, Microsoft’s chief eco-nomic witness, did state in his direct testimonythat the relationship between numbers of applica-tions and sales of operating systems is less clearthan the government claimed. He pointed out thatthe Palm operating system had 1,000 applicationsat the time of his testimony but more than 2 mil-lion users. Apple had more than 12,000 applica-tions and 12.5 million users. By contrast, Linuxhad only 250 applications but 7.5 million users.BeOS had over 900 applications but only 750,000users. United States v. Microsoft Corporation, RichardL. Schmalensee, Direct Testimony, filed January 3,1999, pp. 53–54, http://www.microsoft.com/press-pass/trial/schmal/schmal.asp.

The author has searched for studies that showthere are 70,000 applications. Microsoft does nothave such a study. Interestingly, even Microsoftsupporters have used the 70,000 count to bolstertheir case for Microsoft. For example, syndicatedcolumnist Robert Samuelson wrote, “The worldmight be better off if there were a universal com-puter language that allowed applications soft-ware to run on any machine. Microsoft’s rivals arepursuing such a solution. If they succeed,Windows’ position would decline. Until then,standardization around Windows creates bene-fits. Roughly 70,000 applications are written forWindows, compared with 12,000 for Macintosh.”Robert Samuelson, “Consumers Don’t BelieveThey’ve Been Hurt by Microsoft,” Chicago Tribune,November 19, 1999, p. 31. Similarly, economistNicolas Economides concluded: “[D]espite JudgeJackson’s best efforts, any breakup is unlikely toincrease competition. Justice’s original two-waybreakup plan was premised on the hope that anautonomous applications company would createa new operating system to compete withWindows. But more than 70,000 applications runWindows, creating what the government calls ‘theapplications barrier to entry’ in the operating-sys-tem market. However capable the new applica-tions company, it still wouldn’t be able to single-handedly create a successful rival operating sys-tem. Separately, even with a new applicationscompany’s support, Microsoft’s biggest operat-ing-system competitor, Linux, is unlikely tobecome a serious desktop threat to Windows.”Nicolas Economides, “What’s Worse Than TwoBaby Bills? Three,” Wall Street Journal, May 26,2000, p. A22.

22. Microsoft Corporation, “Windows NTWorkstation Licenses Exceed 11 Million: Top 18

18

PC Manufacturers Demonstrate InnovativeWindows NT Workstation-Based Solutions in theMicrosoft Partner Pavilion at COMDEX,” Pressrelease, Las Vegas, November 17, 1997,http://www.microsoft.com/presspass/press/1997/Nov97/ NTW40Pr.asp.

23. Perhaps a portion of this applications entrybarrier could be surmounted by software. Anapplication is much like “data” that operatingsystems use to perform various functions, likeword processing. That is to say, applications are tooperating systems what “data files” are to applica-tions. That means a new operating-systementrant doesn’t necessarily have to write a totallynew collection of all the applications its systemneeds to be a viable competitor. The new entrantcan play off the installed base of, say, Windowsapplications by writing another application(called an “emulator”) that is compatible with thenew operating system and can interpret Windowsapplications for the new operating system. Thebig disadvantage of that solution is that it intro-duces another layer of code in the form of anoth-er application, which means that the speed of thecomputer system could be impaired (without anoffset in the form of faster computing hardware)and the prospect of errors could increase. Still,computer users could tolerate those problems iftheir only other option was to use Windows thatcarried monopoly prices.

24. No attempt was made to develop a count ofthe unique nonproductivity applications becauseof the difficulty of deciding which applicationsshould be counted.

25. See the original complaint against Microsoftfiled by the government in 1998 (United States v.Microsoft Corporation, Joel I. Klein et al., Complaint,May 20, 1998) and the testimony of the govern-ment’s chief economic witnesses (United States v.Microsoft Corporation, Franklin M. Fisher, DirectTestimony, filed October 14, 1998,http://www.usdoj.gov/atr/cases/f2000/2057.pdf,http://www.usdoj.gov/atr/cases/f2000/2079.htm#footbody_27).

26. There were 476 MBA students surveyed with152 responding, for a more-than-respectableresponse rate of 32 percent.

27. See tables 2.2 and 2.3, along with figure 2.1, inRichard B. McKenzie, Trust on Trial: How theMicrosoft Case Is Reframing the Rules of Competition(Boston: Perseus, 2000), pp. 40, 42. Productenhancements beginning with Windows 3.0 canbe found in table 2.4, p. 43.

28. Findings of Fact, ¶ 408.

29. Ibid., ¶ 409.

30. United States v. Microsoft Corporation,Conclusions of Law, April 3, 2000, p. 22 (emphasisadded). Cited hereinafter as Conclusions of Law.

31. Findings of Fact, ¶ 408.

32. Judge Jackson proposes that Microsoft be splitinto two companies, mainly to encourage the newapplications company to develop a version ofOffice Suite for Linux. It would appear fromApple’s experience that an operating-system firm’shaving its own version of Office is not necessarilythe key to success. The fact that Microsoft haswritten a version of Office for the Mac operatingsystem indicates that Microsoft is not unalterablyopposed to rewriting Office for other operatingsystems when they gain sufficient market share.

33. The judge admits in his findings of fact that “itis not possible with the available data to determinewith any level of confidence whether the price thata profit-maximizing firm with monopoly powerwould charge for Windows 98 comports with theprice that Microsoft actually charges. Even if itcould be determined that Microsoft charges lessthan the profit-maximizing monopoly price,though, that would not be probative of a lack ofmonopoly power, for Microsoft could be chargingwhat seems like a low short-term price in order tomaximize its profits in the future for reasons unre-lated to underselling any incipient competitors.For instance, Microsoft could be stimulating thegrowth of the market for Intel-compatible PCoperating systems by keeping the price ofWindows low today. Given the size and stability ofits market share, Microsoft stands to reap almostall of the future rewards if there are yet more con-sumers of Intel-compatible PC operating systems.By pricing low relative to the short-run profit-maximizing price, thereby focusing on attractingnew users to the Windows platform, Microsoftwould also intensify the positive network effectsthat add to the impenetrability of the applicationsbarrier to entry.” Findings of Fact, ¶ 65. In short,even if Microsoft is not charging monopoly pricesnow, it is still in violation of the antitrust lawsbecause there is a “dangerous probability” that itwill charge monopoly prices in the future. Thejudge concluded: “Even if the first two elements ofthe offense are met, however, a defendant may notbe held liable for attempted monopolizationabsent proof that its anticompetitive conduct cre-ated a dangerous probability of achieving theobjective of monopoly power in a relevant market.The evidence supports the conclusion thatMicrosoft’s actions did pose such a danger.”Conclusions of Law, p. 23 (citation omitted).

34. More precisely, Judge Jackson made this point

19

several times in his findings of fact: “To the extentthat developers begin writing attractive applica-tions that rely solely on servers or middlewareinstead of PC operating systems, the applicationsbarrier to entry could erode. As the Court findsabove, however, it remains to be seen whetherserver- or middleware-based development willflourish at all. Even if such development werealready flourishing, it would be several yearsbefore the applications barrier eroded enough toclear the way for the relatively rapid emergence ofa viable alternative to incumbent Intel-compati-ble PC operating systems. It is highly unlikely,then, that a firm not already marketing an Intel-compatible PC operating system could beginmarketing one that would, in less than a fewyears, present a significant percentage of con-sumers with a viable alternative to incumbents.”Findings of Fact, ¶ 32.

35. Judge Jackson explains “network effects” inthis way: “Consumer demand for Windows enjoyspositive network effects. A positive network effectis a phenomenon by which the attractiveness of aproduct increases with the number of peopleusing it. The fact that there is a multitude of peo-ple using Windows makes the product moreattractive to consumers. The large installed baseattracts corporate customers who want to use anoperating system that new employees are alreadylikely to know how to use, and it attracts academ-ic consumers who want to use software that willallow them to share files easily with colleagues atother institutions. The main reason that demandfor Windows experiences positive network effects,however, is that the size of Windows’ installedbase impels ISVs to write applications first andforemost to Windows, thereby ensuring a largebody of applications from which consumers canchoose. The large body of applications thus rein-forces demand for Windows, augmentingMicrosoft’s dominant position and thereby per-petuating ISV incentives to write applicationsprincipally for Windows. This self-reinforcingcycle is often referred to as a ‘positive feedbackloop.’” Ibid., ¶ 39.

36. Calculating how many consumers useNavigator is difficult because many users buy theircomputers already loaded with both Navigatorand Internet Explorer. But a rough estimate can bemade. There are currently more than 300 millionWeb users in the world. See Marc Saltzman,“Searching on a Sea of Web Pages: SpecializedSites Evolve to Better Help Surfers Find Bearings,”USA Today, January 13, 1999, p. 4D. If Navigator isused by only a quarter of those users—the low endof the commonly reported Netscape marketshare—there would be 75 million or moreNavigator users worldwide.

37. Judge Jackson suggested that Microsoft mighthave charged a monopoly price for its operatingsystem, but he based his argument on nothingmore than an internal study stating that aWindows 98 upgrade could profitably have beensold for $49 rather than its actual price of $89.Findings of Fact, ¶ 63 (“[I]t is indicative of monop-oly power that Microsoft felt that it had substantialdiscretion in setting the price of its Windows 98upgrade product. . . . A Microsoft study fromNovember 1997 reveals that the company couldhave charged $49 for an upgrade to Windows 98—there is no reason to believe that the $49 pricewould have been unprofitable—but the study iden-tifies $89 as the revenue-maximizing price”). JudgeJackson must have realized that the internal studywas a thin reed on which to rest such an inference;he omitted any mention of monopoly price goug-ing from his conclusions of law.

38. See Stan J. Liebowitz and Stephen E. Margolis,Winners, Losers and Microsoft: Competition andAntitrust in High Technology (Oakland, Calif.:Independent Institute, 1999), chaps. 7–9.

39. There is actually a strong argument for allow-ing firms in markets for network goods to“exploit” consumers who are “locked in” becauseof “switching costs.” The ability of a firm toengage in such “exploitation” can lead to lowerup-front prices for consumers and a greater vari-ety of more sophisticated applications. That argu-ment is detailed in Dwight R. Lee and Richard B.McKenzie, “The Case for Letting Firms ExploitNetwork Effects and ‘Locked-In’ Customers,”Graduate School of Management, University ofCalifornia, Irvine, working paper, 2000.

40. Conclusions of Law, p. 8.

41. See Bill Gates, “The Internet Tidal Wave,”memorandum, May 26, 1995, the JusticeDepartment’s exhibit 20, p. 1, http://www.usdoj.gov/atr/cases/ms_exhibits.htm. In that memo to hisexecutive staff, Gates radically upgraded compa-ny-wide interest in and commitment to theInternet, outlining both the threats and theopportunities the Internet presented. He insistedthat his executive staff give the Internet the “high-est level of importance.” Indeed, in that memo, asin his 1995 book, Gates equated the importanceof the Internet to Microsoft with the develop-ment of the first IBM PC and asked all companydivisions to rethink and redesign their productswith the Internet in mind, indicating the prob-lems Microsoft would face in playing catch-up toexisting Internet players and noting that the pro-posed network computer was a “scary possibility.”Later in 1995 in an Internet strategy workshop,Gates once again insisted that the company hadto, in effect, turn on the proverbial dime to make

20

the company’s operating system and applicationsInternet ready. Bill Gates, “Internet StrategyWorkshop Keynote,” internal Microsoft memo-randum, Decembser 7, 1995.

42. The problem Microsoft faces—how radically itcan and should upgrade its operating system todiscourage potential competitors—is one sharedby many successful firms wedded to proven tech-

nologies that they have efficiently exploited for along time. See Michael L. Tushman and CharlesA. O’Reiley III, “The Ambidextrous Organization:Managing Evolutionary and RevolutionaryChange,” California Management Review 38, no. 4(Summer 1996): 1–23; and Clayton MChristensen, The Innovator’s Dilemma: When NewTechnologies Cause Great Firms to Fail (Cambridge,Mass.: Harvard Business School Press, 1997).

21

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