Why Does Microsoft Charge so Little for Windows?
By Bernard J. Reddy, David S. Evans, and Albert L. Nichols
January 7, 1999
Based solely on market share, Microsoft appears to have a near-monopoly in operating systems for personal computers. Yet Microsofts price for Windows is a small fraction-less than 5 percent-of the price that a true monopolist would charge. Moreover, just a few years ago consumers had to spend far more to obtain software that provided just a fraction of the capabilities that Windows offers today. In 1989, consumers would have had to spend more than $250 (in todays dollars) to obtain a fraction of the capabilities they get for $50 with Windows
98 installed on a new computer. Microsoft keeps its prices low and keeps improving its product-despite its high market share-because it faces intense competition from firms that would like their products to displace Windows. Some of its rivals are well-known and well-funded companies, such as IBM, Sun and Apple. But its most threatening competitors are the unknowns-rivals that could arise from the vast pools of highly mobile and talented programmers and venture capital eager to back the next software superstar if Microsoft lets its guard down. The rapid growth of Linux-initially written by a college student in Finland-is just one example of this constant threat.
Some observers have argued that Microsoft Windows is such an important component of most personal computers sold in the market today that computer manufacturers have no choice but to install Windows on their computers. In 1997, about 88
percent of new personal computers based on Intel x86-compatible microprocessors had a version of Windows installed; Intel x86-compatible computers in turn accounted for about 95
percent of all new personal computers. These commentators claim that because Microsoft provides such an important component for personal computers, it is able to charge
prices for Windows. They conclude that the price of Windows is
based on two facts: it costs little to produce and distribute copies of Windows; and the price of Windows has not fallen as rapidly as have the prices of personal computers or computer chips.
The U.S. Department of Justice (DOJ) and 20 states (plus the District of Columbia) made the same claim in an antitrust case filed in May 1998. Using the language of antitrust, they say that Microsoft has
"monopoly power in the operating system market"
which is defined as
"the market for personal computer operating systems that are compatible with Intel x86/Pentium (or Intel-compatible) microprocessors." Professor David Sibley, an economist who was scheduled to testify for DOJ at the trial, prepared a declaration and a later expert report that describe the basis for these conclusions in more detail. In part, he suggests that Microsoft is a monopolist because it could raise the price of Windows-which he says is currently in the range of $45 to $65 -by at least 10 percent and make more money.
Professor Sibley and other critics of Microsoft have actually missed the most intriguing puzzle: if a hypothetical monopolist could profitably raise its operating system price by 10
percent above the levels that Microsoft charges, and if Microsoft has monopoly power in this market, then why doesnt Microsoft raise its price? Furthermore, why would a company that has an important input-the operating system-into a product used widely both in the home and the workplace-the personal computer-charge only about $50 for it. That is only 2.4
percent of the average price of a computer sold in 1997 ($2,082).
Those who claim that Microsoft has monopoly power have missed another paradox. The prices of Microsofts operating systems have risen far less rapidly than the value of its operating systems. Someone in August, 1989 who already had MS-DOS running on a computer would have had to spend about $254 (in 1998 dollars) on other products to obtain just a portion of the software features that are available in Windows 98. And many features that are standard parts of Windows today simply were not available at any price in 1989.
This paper shows that Microsoft charges low prices for Windows and does not increase the price of Windows at the same rate as it enhances the capabilities of Windows for a very simple reason: competition. It is not the static competition that your college economics professor might have used to explain why wheat farmers cannot charge more for an essential food item. Rather, it is the dynamic competition noted by Joseph Schumpeter and described in the modern economics literature on innovation races and superstars. Section II shows that Microsoft acts as if it faces competition, not as if it were a monopolist, in setting the price of Windows. Section III documents how relentless technological progress has resulted in a dramatic increase in the value of Microsofts operating systems software relative to its price. Section IV describes the competition that Microsoft has faced in the past and continues to face today.
Microsoft Acts as if It Faces Competition, Not as if It Were a Monopolist
We are going to demonstrate two implications of Microsofts $50 price for Windows in this section. First, if Microsoft really had a monopoly in the operating system market for personal computers, under plausible circumstances it could charge at least $1,000 for Windows. Second, the fact that Microsoft only charges about $50 for Windows implies that it faces a demand curve with an elasticity whose magnitude is at least 13. That is a highly elastic demand curve and indicates that Windows faces intense competition.
Price-Setting by a Monopolist
Even a true monopolist faces some constraints in setting prices-no firm can set an infinitely high price for its product and actually make sales. Prices for real-world transactions are therefore never infinitely high. In deciding how to set the price for its product, a monopolist must weigh the tradeoff between two effects: a higher price will increase the revenues per unit sold; but a higher price will also decrease the number of units sold, because consumer demand will fall, other suppliers will increase their output of (at least partially) competing products, or both. The monopolist maximizes profits when these two effects are balanced: the direct gain in revenues from a higher price would be offset by the lost net revenues due to the drop in volume caused by that higher price. This condition can be summarized in the following form:
where P is the profit-maximizing price chosen by the monopolist, MC is the marginal cost for the monopolist, and E is the price elasticity of demand. This is a standard result in economic theory.
This equation says that a monopolist should set price so that the net margin equals the inverse of the price elasticity of demand. When demand responds sharply to price changes, the elasticity will be high, and the profit-maximizing price must therefore be close to marginal cost. When demand responds weakly to price changes, the elasticity will be small, and the profit-maximizing price can substantially exceed marginal cost.
Under many circumstances, the demand for one product is
from the demand for another product. For example, someone who does not own a computer has no need for an operating system, and everyone who owns a computer needs an operating system. The demand for operating systems therefore depends directly on the demand for computers.
Suppose that every purchaser of a new personal computer wants exactly one operating system installed on the computer: there is no demand for
machines or for machines with multiple operating systems installed; and OEMs pay for (rather than pirate) the operating systems that they install. Suppose further that the supply of personal computers is highly competitive, with many manufacturers striving to satisfy consumer demands by supplying quality products at the lowest possible prices. If so, the price that an OEM pays for an operating system will be passed on to consumers directly in the price of the computer. Under these circumstances, the demand by OEMs for any one operating system will necessarily equal the number of computers that they sell with that operating system installed (one and only one operating system for each computer).
Further, the price elasticity of demand for a given operating system will be a fraction of the price elasticity of demand for computers equipped with that operating system. That fraction will equal the price of the operating system relative to the price of the computer. If a given vendor of operating systems charges a price P to an OEM, who in turn charges a price P for the computer equipped with that operating system, then the operating system accounts for the fraction P /P of the price of the computer. If the operating system price increases by 1
percent, and that price increase is passed on in the computer price, then the computer price will rise by only ( P /P ) of 1
percent. Demand for computers equipped with that operating system therefore would fall by only ( P /P ) times the price elasticity of demand for computers equipped with that operating system. If the price elasticity of demand for personal computers is 2.0, for example, and if the operating system accounts for only 5
percent of the price of the computer, then the price elasticity of demand for the operating system is only 0.1 (0.1=2*0.05).
Is Microsoft Maximizing Monopoly Profit?
Suppose that Microsoft were actually a monopolist-a firm that could determine its own price without regard to other competitors and could prevent entry into its market. How much would it charge for Windows?
In this hypothetical world of monopoly, every personal computer must have a copy of Windows installed. In its complaint against Microsoft, the U.S. Department of Justice states (p.
"PC manufacturers (often referred to as Original Equipment Manufacturers, or OEMS) have no commercially reasonable alternative to Microsoft operating systems for the PCs that they distribute."
On the other hand, it is widely accepted that the OEM business is highly competitive. Hundreds of OEMs license operating systems directly from Microsoft, and thousands of others obtain licenses through distributors. Dataquest identified more than 30 OEMs that shipped at least 20,000 personal computers in 1997. These top firms shipped barely two-thirds of personal computers, while smaller firms accounted for over 30
percent of the total.
To provide a quantitative estimate of the monopoly price of Windows, we need estimates of the price of a new personal computer, the current price to OEMs of licensing Windows, and Microsofts marginal cost of selling another copy of Windows to OEMs:
The average price of a personal computer is $2,000, based on 1997 industry data.
The average price paid by OEMs for licensing Windows
98 is about $50 per unit.
The marginal cost to Microsoft of licensing an additional copy of Windows to an OEM is zero. This is clearly an underestimate, as Microsoft does incur some support and other costs that vary at the margin (in general, however, OEMs are responsible for support). However, that fact reinforces the conclusions we reach, because our assumption of zero marginal cost leads to an underestimate of the true monopoly price.
. Solving for the operating system price leads to the following expression:
With perfect competition among OEMs, the price of personal computers with Windows preinstalled ( P ) will equal the price charged by Microsoft for Windows ( P ) plus the costs of the other components. These other system costs are therefore estimated to be $1,950 (=$2,000-$50).
The price elasticity of demand for the operating system ( E ) will equal the price elasticity of demand for computers ( E ) times the fraction P / P . The profit-maximizing condition presented above leads to the following derivation (substituting into this condition the expressions for the marginal cost, the computer price, and the operating system price elasticity):
where the last expression follows from the assumption that
Thus, the estimated monopoly price depends critically on the demand elasticity for personal computers.
We know of no reliable empirical estimates of the elasticity of demand that have been estimated specifically for personal computers. Brynjolfsson has estimated that the price elasticity of demand for computers as a whole is in the range of 1.0. The price elasticity of demand for personal computers may be somewhat larger due to substitution possibilities between them and other computers. Economists generally classify the demand for a good as
where the dividing line is an elasticity of 1. Estimates of elasticities for other goods show a wide range, depending on the good in question and the data and analytic structure underlying the analysis. The size of the elasticity depends in part on how narrowly the good is defined; we would expect the elasticity of demand for a Toyota Camry, for example, to be much higher than that for automobiles generally, because there are many close substitutes for the Camry. If one believes that Windows is just one brand among many (as the Camry is), however, then Microsoft does not have a meaningful monopoly (just as the Camry does not). Thus, to be consistent with the assumption of monopoly that underlies this analysis, the relevant elasticity is that for the broader industry category, personal computers (or automobiles). Many demand elasticities at the industry level are estimated to be less than 1.0, and very few exceed 2.0. For example, the demand for automobiles has an estimated elasticity of demand in the range of 0.6 to 1.2. Another recent study of the demand elasticity for consumer durables found a median elasticity of demand of 1.4.
We conclude that a price elasticity in the range of 1.0 to 2.0 is plausible for personal computers, with the lower end of this range agreeing with Brynjolfssons 1.0 for computers as a whole. Thus, for most of our calculations we use 2.0 as our value for E , which is conservative in the sense that it gives a
estimate of the monopoly price.
With a demand elasticity of 2.0 for personal computers, the optimal monopoly price for the operating system is:
If the elasticity of demand is 1.5, which is still elastic by normal criteria, the monopoly price doubles, to $3,900.
An operating system price in excess of $1,000 may seem high given that most copies of Windows
98 licensed through OEMs are priced at far less than $100, but such prices are not at all unheard of in the computer software business. For example, Lotus SmartSuite and Corel WordPerfect Suite, office suites that include a spreadsheet, a word processor, and other applications, sell at retail for prices in the range of $300-$400, despite facing competition from each other and from Microsoft Office. Stata, a widely used statistical program, sells for $965 for a single copy, despite competing with products such as SAS, SPSS, and a variety of other, perhaps more specialized, products (EViews, TSP, LIMDEP, Shazam, S-Plus, Gauss, and other statistical or econometrics packages). AutoCAD, a professional drafting program, has a suggested retail price of $3,750. Prices for other commercial operating systems that run on Intel-compatible hardware are frequently in the range of $250 or more per copy.
Can Complementary Products Explain Low Prices for Windows?
In the real world, additional considerations also affect the pricing of products like operating systems. Microsoft offers many products that are complements to its consumer versions of Windows, such as Microsoft Office (word processing software, spreadsheet software, and other office productivity applications), Microsoft Encarta (multimedia encyclopedia), and so forth. If Microsoft succeeds in promoting one additional sale of a computer with Windows
98, it has some likelihood of making additional sales of complementary products such as these, whether through OEMs, retail suppliers, or other channels. Of course, additional sales of complementary products are not guaranteed: not all computer users need the particular complements that Microsoft offers; and all of these complementary Microsoft products face substantial competition from products from other software vendors (e.g., Corel and Lotus, among others)-even restricting attention to competing products that still can run on Windows.
Because of the possibility of earning revenues from these complements, a monopolist over the operating system would have an incentive to temper its operating system prices. In contrast, a perfectly competitive supplier of a stand-alone operating system (a vendor of operating systems but not complementary products) would not have that incentive to keep prices low. The existence of complements can have noticeable impacts on pricing.
Suppose that net revenues (from complementary applications) equal to R are available for each unit of an operating system licensed through an OEM. The profit-maximizing condition presented above leads to the following derivation:
Solving this for the profit-maximizing price leads to the following expression:
According to Microsofts 1997 annual report, in fiscal 1997 the firms gross revenues from its applications group ($5.39
billion) were approximately equal to its gross revenues from its platforms group ($5.97
billion), which includes operating systems. If the
group as a whole is considered a complement to the
group as a whole, then (on average) Microsofts complementary gross revenues per OEM unit of Windows are similar in magnitude to its average price for Windows
98. Other variations are possible, but it seems highly unlikely that complementary gross revenues per OEM unit of Windows
98 are more than a small multiple of Microsofts average price for Windows
Suppose that Microsoft could expect to earn complementary net revenues of about $100 for each unit of Windows
98 that it licenses through OEMs; this figure is roughly double the average OEM price we are using ($50 per unit) and therefore corresponds to gross revenues that are more than double the OEM price of Windows. Then with a price elasticity for personal computers of 2.0, the profit-maximizing price is $1,750 instead of $1,950. Thus, the availability of revenues from sales of complementary products still cannot explain the gap between Microsofts actual OEM prices for Windows and what the short-run profit-maximizing price would appear to be-even if the complementary net revenues are several times as large as we have used in these calculations.
Microsofts Windows Prices Are a Small Fraction of the Price a Monopolist Would Charge
Even if we account for sales of complementary products, the price actually charged by Microsoft for Windows is more than an order of magnitude lower than a firm would charge if it had the secure monopoly that many observers (including the Department of Justice) appear to believe it has. Table 1 summarizes the prices that a monopolist would charge for Windows under several assumptions about the elasticity of demand for personal computers and the importance of complementary sales. It reports these figures for two different assumptions about the demand for personal computers-a constant elasticity of demand schedule (so that the demand elasticity is constant at all prices) and a semi-log demand schedule (so that demand becomes more elastic at higher prices). The monopoly price ranges from a low of $900 (semi-log demand, elasticity=2, complementary net revenues = $100 per unit of Windows) to a high of $7,800 (constant elasticity demand, elasticity=1.25, zero complementary net revenues).
Table 1. Monopoly Operating System Price Depends on Elasticity of Computer Demand, Complementary Net Revenues, and Shope of Demand Curve
Constant Elasticity Demand
Note:R is the net revenue from complements per unit of operating system; for the semi-logarithmic demand, the elasticities are assumed at a computer price of $2,000 and increase as the computer price gets higher.
Source:As noted in text.
Microsofts Prices Are Consistent With Its Facing a High Elasticity of Demand
If one had to place the $50 price of Windows somewhere along the spectrum of prices that could emerge from the usual static economic models, it would fall far closer to the price that would emerge from perfect competition (roughly marginal costs plus average development costs plus a return on those costs that compensates for the fact that most software investments do not pan out) than the price that would emerge from perfect monopoly (roughly $1,000 or more). As discussed in more detail in Section IV, we believe that the answer to this apparent inconsistency-a
who charges relatively low prices-is that Microsoft in fact faces a great deal of competition, from existing firms and from potential new entrants. Clearly it does not make sense to think of the demand elasticity for Windows as being derived entirely from the demand elasticity for personal computers. Microsoft prices Windows as low as it does because of several types of substitution, not just the substitution of other devices for personal computers: the substitution of personal computers that are not Intel-compatible for those that are; the substitution of other operating systems for Windows on Intel-compatible computers; continued use of existing computers (and their operating systems) by users, foregoing the purchase of new units (with new operating systems); and piracy of Microsofts operating systems.
Economists often use the elasticity of demand as a proxy for how much competition a firm faces. A higher elasticity of demand indicates that consumers have more substitution possibilities and that the firms ability to raise prices is limited. It is possible to take the mathematical apparatus presented above and use it to answer the following question: at what elasticity of demand is the profit-maximizing price for Windows $50. With a constant-elasticity demand schedule, the demand elasticity equals:
where NR equals the average net revenues available from complements for every copy of Windows that Microsoft licenses. Inserting $50 for P and $2,000 for P yields an elasticity of demand of 40 when net revenues from complements are not available. Assuming that net revenues from complementary products are $100 per unit yields an elasticity of demand of 13. The elasticity of demand for Windows is similar to that for many narrowly defined brands for which there are many substitutes. For example, Tay and McCarthy found that price elasticities for particular types of vehicles ranged from 0.5 to 6.1; Berry, Levinsohn, and Pakes found price elasticities for individual makes of cars in the range of about 3.1 to 6.8. It therefore appears that Microsoft sets its prices as if it believes that it faces the kind of highly elastic demand curve that characterizes firms that produce products for which there are many readily available substitutes.
The Quality-Adjusted Real Price of Windows Capabilities Has Fallen
Microsoft therefore charges a
price for its operating system relative to the monopoly price. Moreover, its price is
taking into account the features that have been added to Windows. When Windows
1.0 was released in 1985, it was a limited operating environment that ran in conjunction with an operating system such as MS-DOS or its IBM-labeled twin, PC-DOS. Both Windows and MS-DOS were improved over the years through 1993, when Windows
3.11 and MS-DOS
6.2 were released; these were the last major products in the Windows
3.x and MS-DOS lines. Windows
95, released to great fanfare in August 1995, essentially combined the operating system underpinnings formerly provided by MS-DOS (and enhanced by Windows
3.x) with the graphical user interface formerly provided by Windows
3.x into a unified whole, with both the operating system and the GUI greatly improved. In this regard, Windows
95 became like other operating systems such as the Apple Macintosh operating system, IBMs OS/2 (for Intel-compatible hardware), and Microsofts own Windows
95 went through several interim upgrades (installed by OEMs on new computers or (for most components) available as free downloads directly from Microsoft) and has been replaced by Windows
We compare the current OEM price for Windows 98 with the price that consumers would have had to pay in 1989 (before Windows
3.0 was released) for a fraction of the capabilities currently provided by Windows
98. This comparison is complicated by several factors:
Until August 1995, a copy of MS-DOS (or a clone of MS-DOS) was needed to boot a system in order to run Windows. Windows
95 and its successor, Windows
98, are stand-alone operating systems. As a result, comparisons over time must include the price of the companion booting operating system, when relevant.
Microsoft at various times has licensed new copies of Windows and MS-DOS to OEMs for installation on new computers, upgrades to end users for older versions of Windows and MS-DOS, and (at least for Windows) new copies (not upgrades) for end-users who did not have an older version of Windows. In general, however, Microsoft has not licensed full MS-DOS products at retail to end users.
Prices for OEM products are generally not available.
Retail prices for various products vary in their availability.
There is no obvious way to control for changes in product quality over time.
We deal with each of these problems in the following analysis.
Changes in Product Quality
The first popular version of Windows was 3.0, released in May 1990, typically running in conjunction with MS-DOS 4. When Windows
3.1 was released in April 1992, it typically ran in conjunction with MS-DOS
5. Versions of Windows for Workgroups (3.1 and 3.11) and MS-DOS (6.0, 6.2, 6.21, and 6.22) were released between October 1992 and June 1994. Windows for Workgroups was a full-fledged member of the Windows
3.x family of products, but it was priced slightly higher than Windows
3.1 and provided peer-to-peer networking. The staggered releases of MS-DOS and Windows
3.x provided an opportunity for Microsoft to ship improved MS-DOS components that were useful for Windows, such as disk caching software and memory managers. Windows
95 simultaneously replaced MS-DOS and Windows
3.x when it was released in August 1995. Windows
98 replaced Windows
95 in June 1998.
Important release dates and notable new capabilities of these products are summarized in Table 2. In general, capabilities that were introduced in one version of MS-DOS or Windows were retained in future versions, unless they were made obsolete.
Table 2. Release Dates and Summary of New Features of Major Versions of MS-DOS and Windows, 1990-1998
Source: Various product manuals, reviews in the trade press, and Computer Desktop Encyclopedia .
In some cases, the new capabilities of a given release of MS-DOS or Windows could have been provided with stand-alone utilities that were previously available. For example, PC
Tools for DOS was a collection of utilities that, as of 1989, provided menu capabilities, file undelete, disk unformat, disk caching, a DOS shell for file management, disk backup, disk defragmenting, and communications; later versions added even more utilities. Its
included notepads, calculators, an appointment scheduler, and a cut-and-paste utility that could be used to transfer information from one DOS program to another. Other vendors in 1989 marketed utilities either singly (e.g., PC-KWIK for disk caching) or in combination (e.g., Norton Utilities, which partially overlapped the capabilities of PC
Tools, other than the desktop manager). QEMM was a memory manager available in 1989, when MS-DOS had no such capabilities; later versions of QEMM enhanced the memory management capabilities of the memory managers built into even the last versions MS-DOS and Windows 3.x. Also available in 1989, DESQview was a highly-regarded, multitasking, windowed environment for running DOS-based programs. Various fax programs existed for DOS (such as BitFax); later fax programs for Windows
3.x (such as WinFax) made faxing from within Windows 3.x little more difficult than printing to a standard printer. Stacker was perhaps the best-known of the disk-compression packages, becoming available around the beginning of 1991. TCP/IP software for DOS and Windows 3.x became available from a number of vendors. Lantastic provided peer-to-peer networking for DOS and Windows 3.x computers.
A number of other capabilities, however, could be duplicated only with great difficulty, if at all, before being integrated into Windows. The direct operating system support for printers, fonts, modems, and sound cards introduced in various versions of Windows all greatly simplified the prior state of affairs. Every DOS program that wanted to allow for printing with other than the plainest of text had to provide printer drivers for every popular printer, taking into account the fonts and other capabilities of each printer. The printer drivers installed with Windows 3.x took care of that for Windows-based programs. Much the same was true for communications packages, fax software, remote control utilities, and other programs that needed to control modems; rather than develop drivers for every popular modem, the software vendors can rely on the modem drivers in Windows
95 and Windows
98. The same again is true of games and, increasingly, mainstream business applications that use sound. Before sound cards became directly supported by Windows, these applications had to provide their own drivers for every popular sound card.
TrueType fonts fall into a slightly different category. To some extent, font scaling utilities could be used in Windows 3.0. In practice, however, no application program could rely on the assurance that such a utility would be available on a given computer, or on what fonts would be installed if such a utility were available. The inclusion of TrueType font technology with Windows 3.1 finally guaranteed that a font scaling utility would be available on every Windows
3.1 desktop, that some set of fonts (e.g., Times New Roman, Arial, Symbol, Courier New) would be available on those desktops, and that the fonts on the screen would match the fonts on the printed page. This helped cause an explosion of interest in Windows applications that could make use of these capabilities.
The capabilities of OLE, a method to link information between documents, were impossible to duplicate with DOS or Windows applications before they were supported in Windows 3.x. OLE permitted users to copy information from one Windows application and paste it into another-with a live link that would permit the target document automatically to reflect changes in the source document.
Value to Users of These Product Improvements
There probably do exist users of Windows
95 or Windows
98 who have little or no need for these enhancements to Windows over the years and who would have had no need for related capabilities before they were added to Windows. A person with no need for any of these capabilities, however, would not do any of the following:
connect to a network (peer-to-peer networking, direct network support);
use a modem (direct modem support, communications software, fax software, dial-up-networking);
access the Internet (TCP/IP, FTP, and browsing support);
print formatted documents (direct printer support, scalable fonts);
play modern, multimedia games (direct sound card support, direct video support);
link information between two documents (OLE);
run multiple applications or large single applications (memory managers, task switching and/or multitasking);
perform basic file and hard disk maintenance-including recovering deleted files, defragmenting hard disks, marking bad clusters (undelete, defragmenting, and disk scanning capabilities);
compress a disk to postpone the need to purchase a larger hard disk (disk compression).
In contrast, a substantial number of users of Windows
95 or Windows
98-perhaps even a majority-are likely to need most of these capabilities. Disk compression is perhaps the least important-with rapidly falling prices for hard disks, it may be easier (but not cheaper) to install a second hard disk than to activate the disk compression software in Windows
98. Quite commonly, however, a user in a corporate setting needs many of these capabilities, even sound card support-modern business applications are increasingly using sound. Quite commonly, a small office or home business user would need most of these capabilities. And also quite commonly, a home non-business user would need most of these capabilities (with the various network capabilities being used to connect to the Internet or to office networks).
Table 3. Operating Systems and Utilities Needed in 1989 and 1998 for Similar Capabilities
Capabilities Sources:Various product manuals, reviews in the trade press, and Computer Desktop Encyclopedia .
Trying to develop a year-by-year, quality-adjusted price index for Windows appears to be impossible. It is possible, however, to take some snapshots of the prices for Windows, DOS, and various utilities needed in the past to provide some semblance of the capabilities currently available in Windows
98. Table 3 summarizes the combinations of non-Windows products as of 1989 (before Windows
3.0 was released) needed to provide some of the useful features of Windows
98. In 1989, many important capabilities of Windows
98 were either unavailable (e.g., file linking) or not readily available (e.g., cross-application support for scalable fonts, printers, modems, and sound cards). Some other capabilities would have been either irrelevant (e.g., Web browsing) or less widely useful (e.g., TCP/IP support) in 1989 than in 1998.
In 1989, many of the current capabilities of Windows 98 could have been obtained by using a combination of MS-DOS
5.5, QEMM386, and DESQview, although with much reduced ease of installation and ease of use compared with Windows
98. Table 4 summarizes what these combinations of products would have cost at retail in 1989 and 1998. These comparisons are complicated by the fact that Microsoft never sold MS-DOS at retail, and prior to 1991 (with the release of MS-DOS 5) it never sold MS-DOS upgrades at retail. They are also complicated by the fact that Microsofts OEM prices for MS-DOS and Windows are not publicly available.
Table 4. Prices of Products Needed for Comparable Capabilities: 1989, 1991, and 1998
Product Sources:OEM price for Windows 98 assumed to be $50. All other prices are for mail order purchases from PC Connection, a major mail order vendor for over a decade. August 1989 prices from advertisement in that month's issue of PC Magazine .
Based on an estimated current price of $50 for copies of Windows
98 licensed through OEMs, we see that the current capabilities included in Windows
98 today cost the user far less than did even a small fraction of those capabilities in 1989. Even ignoring inflation, the current $50 price for Windows is less than half the 1989 price for DESQview
386 alone. In August 1989, the combined prices of DESQview
386 and PC
Tools were $194. This total excludes any implicit price for MS-DOS. Public data on average OEM prices for MS-DOS are not available, but the prices were substantially less than the current price of Windows
98. The consumer price index rose approximately 31
percent between August 1989 and August 1998. In order for the older prices to be made comparable with the 1998 prices, the August 1989 figure should be multiplied by 1.31. In 1989, it would have cost consumers $254 (in 1998 dollars), plus the cost of MS-DOS, to obtain a fraction of the capabilities that now cost $50 with Windows
One might argue that only so-called
of personal computers would actually have spent over $200 per computer for utility software in the days before Windows
3.0 became popular (despite the fact that many people did indeed purchase these and other utilities, some similar and others not). If so, then the comparisons presented here might be considered to be biased in favor of Windows
98-they include too much software for 1989. An opposing argument, however, seems more likely to be valid: the non-power users are the ones who gained most from the emergence of Windows. The gains in ease of use, in configuring and using modems, sound cards, and printers, in connecting to the Internet, and in using screen fonts that match printer fonts-these are the gains that probably matter most to people who are not power users. And these gains are not reflected in any way in the price comparisons presented here.
The inescapable conclusion to be drawn from these figures is that the quality-adjusted, real price of obtaining the capabilities currently reflected in Windows
98 has fallen sharply since 1989. This is true for capabilities that were reasonably available as of 1989. It would be even more true if the comparison could reflect capabilities that were not available in 1989 but would have been highly valued had they been available (e.g., printer, modem, sound card, and scalable font support).
Comparing Operating System and Hardware Prices
As mentioned above, some observers have attempted to compare pricing over time for Windows with pricing over time for personal computers or microprocessors. Such comparisons are meaningless. The products have different cost structures and different rates of technological change. There is no more reason for the prices of software and computers to behave similarly over time than there is for the prices of gold and Golden Books to behave similarly.
Microprocessors and personal computer hardware embody tremendous advances in technology for semiconductors and electronics. As a result, there is good reason for the prices of personal computers (holding quality constant) to have fallen markedly over, say, the last decade. Alternatively, there is good reason for a $3,000 computer today to be much more powerful than a $3,000 computer a decade ago.
Writing computer software, whether today or a decade ago, takes time and effort by computer programmers. Technological change in programming is quite different from technological change in hardware. Programming tools have certainly improved over the past decade, but so have user expectations for computer programs. Programs like operating systems have, on average, gotten much larger over the past decade. Whether programming tools have improved programmer productivity faster than user demands have increased program complexity and size is not self-evident. In general, however, there is no reason to believe that the costs of writing an operating system today are less than the costs of writing an operating system a decade ago-todays operating system is much more complex than that of a decade ago. (The marginal costs to Microsoft of licensing an operating system through an OEM are approximately the same today as they were a decade ago: close to zero.)
In general, then, there is no cost-based reason for the price of Windows today to be less than the price of Windows a decade ago. The
of today is far more capable than the
of 1993 and the
of 1988. In contrast, there do exist cost-based reasons for personal computer hardware to have gotten both less expensive and more powerful over time.
Microsoft Has Faced, and Continues to Face, Competition
Microsoft charges low prices for Windows and has steadily improved the quality of its operating system not because it is benevolent: vigorous competition has forced Microsoft to behave this way. Microsoft has faced competition since it released its first operating system in 1981, and it continues to face competition today. This competition has occurred at many levels, involving the speed, power, and ease of use of operating systems. The discussion below concentrates on major products visible to the public. It ignores numerous other products (e.g., OS-9, Theos, CTOS, QNX, and PTS-DOS) that run on Intel-compatible hardware.
Entry in the Past 40
Microsoft has faced operating system competition at all times since its first operating system was released. When IBM started to ship its first personal computer, in 1981, it announced the availability of three operating systems: PC-DOS (written by Microsoft for IBM); CP/M-86 (from Digital Research), and the UCSD p-System (from Softech Microsystems). CP/M-86 had the advantage of being a rewrite of the most popular operating system for Intels 8-bit microprocessors (8080 and 8085) and related CPUs from other vendors (Zilogs Z80). The p-System had the advantage of running on multiple hardware platforms (such as the Apple II and the 8080/8085/Z80 computers that typically could run CP/M). PC-DOS had the advantages of being much cheaper than CP/M-86 ($40 instead of $240), immediately available with applications (CP/M-86 was not immediately available and had no immediately available applications from IBM), and much faster than the p-System. PC-DOS won the early race for popularity on the IBM PC. Of course, the early IBM PC faced competition from the Apple
II and the many brands of computers running CP/M, such as Osborne, Kaypro, Xerox, Radio Shack, Cromemco, IMSAI, Morrow, and many others.
When Microsoft began licensing its operating system to OEMs under its own branded name (MS-DOS), it faced competition not only from CP/M-86 and the p-System but also from its near-twin, PC-DOS. IBM sold its version of the operating system at retail. As a result, consumers could purchase a copy of PC-DOS and run it on a
of the IBM-PC, unless the manufacturer of the clone had customized its hardware so that a generic operating system would not run properly. An OEM who did not like the license terms offered by Microsoft could always tell its customers to pick up a copy of PC-DOS. Even after PC-DOS and MS-DOS stopped being virtually identical (beginning with version 6 in 1993, they had different utilities for memory management, for example), a consumer could purchase a copy of PC-DOS for use on non-IBM computers. Microsofts MS-DOS therefore faced competition from the essentially identical PC-DOS.
In the early to mid-1980s, Microsoft had to weather several other attacks. Digital Research continued to produce new products, such as its Concurrent DOS. Various vendors began porting operating systems from other hardware platforms to the IBM PC. Microsoft itself had developed a 16-bit version of UNIX (Xenix), but other vendors were active as well. Pick ported its operating system from larger computers to the IBM PC.
Competition arose on other fronts as well. Apples Lisa (1983), followed by its Macintosh (1984), introduced a graphical user interface (GUI) to the general public. Several vendors became interested in developing interfaces for the IBM PC that would provide some combination of windowing, multitasking, and a GUI. VisiCorp, publisher of VisiCalc, showed its integrated software system VisiOn at the Fall Comdex show in 1982. This integrated operating system/applications package introduced windows to the IBM-PC world of hardware. Unfortunately, it could not run DOS programs, it was slow and buggy, and it was expensive. In September 1984, Digital Research announced GEM (Graphics Environment Manager), which provided a graphical interface but no multi-tasking; this product made at least some headway when Tandy (Radio Shack) announced that it would offer the product for its computers. Quarterdeck Office Systems developed a highly-regarded product that could multi-task standard DOS programs in windows, although its interface was not graphical. Originally called DESQ, it later became DESQview. In February 1985, IBM released TopView, a multi-tasking program that also provided windows but not graphics. Microsoft announced Windows in November 1983 and finally shipped it in November 1985.
GEM Desktop, TopView, DESQview, and the first release of Windows all differed fundamentally from the failed VisiOn: all were run after a computer had been booted with MS-DOS or PC-DOS. All provided windows for different applications. GEM and Windows provided a graphical interface, while TopView and DESQview did not. DESQview excelled at multi-tasking DOS programs. The other three programs provided a platform for which software developers could write new applications; several GEM applications were available from Digital Research. GEM could run standard DOS applications, but not in a window; nor could it multi-task or switch between DOS applications. In a review of these programs in their infancy, an article in PC Magazine presciently predicted that at most one would prosper:
Consequently, it seems likely that only one of these alternative operating environments will survive. If none of them makes it, then something else will. The Macintosh proved that a command line interface is unnecessary and antiquated. The once-ubiquitous DOS prompt may be a dying breed.
TopView was withdrawn from the market in mid-1987. GEM eventually found a home as the operating environment for Atari computers. DESQview attracted a dedicated group of followers until the capabilities and growing popularity of Windows
3.0 (released in 1990) eliminated the need for a program whose strength was multi-tasking DOS programs. Microsoft kept developing and enhancing its Windows product. At the end of 1987, Microsoft started shipping Windows
2-and its own first high-quality Windows application, Excel for Windows.
The operating system wars continued on other fronts. IBM and Microsoft had cooperated in developing a new operating system, called OS/2. Designed to be more powerful and stable than MS-DOS, it was initially released (without a graphical front end) in December, 1987, but met with little success. The first version of OS/2 with a GUI (1.1) shipped October
31, 1988-but still with relatively poor support for DOS applications and relatively few device drivers to support hardware other than IBMs.
In mid-1988 came DR
DOS, the latest operating system product from Digital Research. This time, however, Digital Research released a product that attempted to be compatible with MS-DOS, much as PC-DOS was but without the advantage of relying on the same underlying operating system code as MS-DOS. It provided some additional utilities and capabilities, but it also attempted to run DOS programs in exactly the same ways that MS-DOS did. Subsequent releases by Digital Research of improved products put pressure on Microsoft to improve MS-DOS, with MS-DOS
5 shipping in 1991 and MS-DOS
6 in 1993. Under the terms of the agreements between IBM and Microsoft, IBM also released new versions of PC-DOS in the same time period.
In May 1990, Microsoft began to ship Windows
3.0, the first truly successful version of the product. Some of the remaining major gaps in capabilities were addressed two years later with the release of Windows
3.1, which provided true what-you-see-is-what-you-get capabilities to Windows applications (along with other features). These improvements finally moved Windows ahead of a host of other products that for the most part merely ran on top of DOS: GEOS, a graphical windowing environment running on top of DOS, released to substantial praise in late 1990; various multi-tasking environments, including DESQview
386, VM/386 MultiTasker, Omniview
386, and Vmos/3; and task-switching programs such Software Carousel, HeadRoom, and other products.
About the time that OS/2
1.1 shipped in late 1988, Microsoft began work on its own advanced operating system-what eventually became Windows
NT. In late 1990, the Joint Development Agreement between IBM and Microsoft, which had led to the development of OS/2 (among other products) unraveled. Microsoft emerged with control of the NT work, while IBM emerged with OS/2. The first version of Windows
NT finally was released in July, 1993-more than a year after OS/2
2.0, which finally won praise for support of DOS and Windows applications. IBM promoted this version of OS/2 as a
"Better DOS than DOS"
Better Windows than Windows. & quote;
After its agreement with Microsoft was dissolved, IBM formed an alliance (Taligent) in 1991 with Apple to develop an object-oriented operating system (based on Apples Pink) that would run on future computers to be built by each firm. That partnership was essentially dissolved when Taligent became an IBM subsidiary in 1996, with no operating system emerging from the joint venture.
Meanwhile, UNIX had not disappeared; to the contrary, it was thriving. In 1988, NeXT introduced its computer running a UNIX-based operating system with a GUI, NeXTStep. When NeXT later gave up on the hardware business, it ported its operating system to other platforms, including Intel-compatible computers. NeXT has since been purchased by Apple; originally, Apple expected NeXTStep to form the basis for future operating systems for Apples computers, but those plans have since changed. Also in 1988, the Open Software Foundation was founded by a consortium of seven companies, including IBM, to promote standards in the fractured world of UNIX. Work undertaken by the OSF eventually led to Motif, a standard GUI for use with different versions of UNIX. This work built on the X-Windows development performed earlier at MIT to produce a windowing environment for UNIX. Several vendors (such as SCO and Sun) currently sell versions of UNIX that run on Intel-compatible computers, and yet more versions (e.g., from Sun, IBM, Hewlett-Packard, and other firms) run on other computers that compete with Intel-compatible computers to various degrees.
Over the last five years, Microsoft has continued to release improved versions of its operating systems. Major milestones were the releases of Windows
95 (August 1995), Windows
4.0 (July 1996), and Windows
98 (June 1998). IBM has also continued to improve OS/2, with major new releases in 1994 (version 3) and 1996 (version 4).
In summary, Microsofts operating systems for Intel-compatible computers have faced competition since they were first released. This competition has occurred at various levels: from operating systems that were not compatible (e.g., CP/M-86, UNIX); from alternate operating environments (e.g., TopView, DESQview, GEOS, GEM) that ran on DOS; from compatible operating systems (e.g., DR
DOS); from advanced, compatible operating systems with a GUI (e.g., OS/2
2.0 and later); and from other hardware platforms that compete with Intel-compatible computers (e.g., the Macintosh and PowerPC, NeXT). And Microsofts MS-DOS had to compete against a near-twin: PC-DOS.
Microsoft faces competition from existing operating systems that run on Intel-compatible hardware (and often other hardware as well), such as OS/2, BeOS, Linux, Unixware, Solaris, and so forth. Microsoft also faces competition from operating systems running on other hardware, such as the Macintosh and proprietary hardware from other firms. Besides the Macintosh operating system, this would include BeOS (running on Apple hardware), Linux (running on various hardware), and Solaris (running on Suns hardware), as well as other proprietary versions of UNIX running on proprietary hardware. Microsoft also faces competition from new and emerging technologies, some of which are currently visible and others of which are not as yet.
The flurry of interest over the last year or two in the"
an example of a new or emerging technology, has posed a threat to Microsoft, at least for some computer uses. Objectives for the network computer included low initial cost and ease of administration-with standardized hardware, no local storage (e.g., no hard or floppy disks), a simple operating system not based on Windows, and a browser as a user interface. Widespread acceptance of the network computer could have substantially cut into sales of personal computers running Windows. To date, the network computer has not succeeded in capturing large numbers of desktops. It did, however, encourage computer manufacturers to release low-cost computers and Microsoft to reduce the administration costs (initial setup, software maintenance, and so forth) for networked computers running Windows.
Currently Linux and the BeOS may provide the biggest near-term competition to Microsoft, although the threat of the network computer has not disappeared. Both are currently available for Intel-compatible hardware (and other hardware as well). Both are Posix-compliant, with Linux essentially being a flavor of UNIX and the BeOS sometimes described as UNIX-like. Linux is a freeware product that is available (with technical support and various utilities and applications) from commercial vendors such as Red Hat and Caldera. Linux is already popular for Web servers and file servers. Commercial software firms (such as Netscape, Oracle and Informix) have recently announced plans to port applications to Linux. If desktop applications follow (such as Corels announced port of WordPerfect Suite), Linux could displace Windows on a substantial number of both servers and desktops. The BeOS is a commercial product developed and marketed by Be, Inc., with initial applications targeted at video and audio processing involving multiple processing threads and huge file sizes. If its file handling proves superior to that of other operating systems, it could become a desirable platform for other applications software as well.
The Palm OS, EPOC, and similar operating systems for handheld computers currently provide a different, perhaps more distant, type of competition for Microsofts desktop operating systems. The Palm OS is used in 3Coms PalmPilot series of handheld computers, while EPOC is used in Psions line. Handheld computers with these operating systems currently compete directly with other models running Microsofts Windows CE; product reviews to date have generally rated the Palm and Psion products as superior to those running Windows CE. At present, none of these operating systems (including Windows CE) will run on a desktop (or laptop) computer. In principle, however, there is no reason why the vendors of these products could not do the reverse of what Microsoft has attempted: Microsoft has tried to write a scaled-down, Windows-style operating system that can run within the current hardware limits of the handheld personal computers; these other operating systems could in principle be scaled up to run on low-cost desktop personal computers. Psion has recently reached agreements with major cellular phone manufacturers (Ericsson, Nokia, and Motorola) for a joint venture (Symbian) to develop and market versions of its operating system for devices with cellular phone capabilities. Microsoft has announced that hardware vendors are developing the AutoPC, a computer running Windows CE that would handle various tasks in automobiles. And Microsofts Windows CE
2.1 will run on a new class of mini-notebook computers that are larger than the current class of handheld computers. Given that these operating systems already are being ported to various hardware platforms (including the movement of CE to the mini-notebook), it seems plausible that they could be ported to desktop computers as well.
Based solely on market share, it would be easy to conclude that Microsoft has a near-monopoly in operating systems for personal computers. Yet, as we have shown, Microsofts price for Windows is a small fraction of what a monopolist would charge. Furthermore, that price is much less than what consumers would have had to pay in the past for even a fraction of the current capabilities of Windows. The answer to this apparent puzzle is that although Microsoft has what would appear to be great market power based on traditional static measures of competition, it has faced, and continues to face, substantial dynamic competition from a host of competitors or potential competitors. Some of these competitors (e.g., IBM, Sun) currently sell products that run on the same hardware as Microsofts products. Others sell products that run on other hardware (e.g., Sun, 3Com, Apple). Others could arise from the vast pools of highly mobile and talented programmers and venture capital eager to back the next software superstar. Entry into operating systems for personal computers has occurred repeatedly over the two decades of the industrys existence. There is no reason to believe that entry has suddenly come to a halt.
In typical manufacturing industries, firms with high market shares arise because of high barriers to entry. They own key patents, have exclusive access to essential natural resources, own highly specialized and expensive physical capital that has large economies of scale, or have government-granted exclusive franchises. Microsoft has none of these barriers working in its favor. In the short run, it benefits from
Users care not only about the quality of the operating system and its price but also the availability of a rich set of applications software that will run on the operating system. These effects mean that at any time, one operating system is likely to capture most of the market. That position, however, is not protected in the
run, which in the computer industry is measured in months or a few years, not the decades that may be required for capital turnover in other industries.
The market for operating systems is best seen as an overlapping series of races or contests to become the
The odds against any one contestant are long, but the rewards are sufficiently great-and the factors of production sufficiently mobile and readily supplied-that a steady stream of contestants is guaranteed. The firm that rests on its success and fails to innovate will quickly be overtaken by other firms that seize the opportunity to take advantage of advances in hardware that open up new possibilities. The firm that tries to exploit its ostensible monopoly through high prices will quickly find that lower-price clones claim the market, or that users will switch to a new platform with its own set of applications. Only through a combination of frequent innovation and relatively low prices can the leader in such a race maintain its position.
AccessMicro, http://www.accessmicro.com, visited 5 August 1998.
"Intelligent Life Discovered at IBM: OS/2 to be Bundled," Infoworld , 4 May 1992, p. 4.
Anthony, Robert S.,
"The Ultimate Personal Peripheral," PC Magazine , 10 March 1998.
"Stacker 3.1 for Windows & DOS," PC Magazine, 14 September 1993, p.
"Be Product Price List,"
1998, http://www.be.com/purchase/price_lists/index.html, visited 6 August 1998.
"The BeOS Features: Is BeOS Similar to Unix?"
1998, http://www.be.com/support/qandas/faqs/faq-0152.html, visited 6 August 1998.
"The BeOS Features: Is BeOS Unix-Based (or Mach-Based)?"
1998, http://www.be.com/support/qandas/faqs/faq-0131.html, visited 6 August 1998.
"The Be Operating System,"
1998, http://www.be.com/products/beos/beos_ds.html, visited 6 August 1998.
Berry, Steven, Levinsohn, James and Pakes, Ariel,
"Automobile Prices in Market Equilibrium," Econometrica , July 1995, pp. 841-890.
"WinFax Pro 3.0: sophisticated fax software," PC World, January 1993, p. 86.
"Microsoft Keeps Pricing Steady For Its Windows 95 Software," The Wall Street Journal Interactive Edition , 23 March 1998.
"Software Platforms Lead to New Products," Electronic News , 19 January 1998, p. 58.
Brousell, David R.,
"MS-DOS, CP-M, Unix Battle for Domination in Market Demanding Multi-tasking, Multi-User Capability," Electronic News , 26 September 1983.
"The Handheld that Beams," PC Magazine , 5 May 1998.
"Some Estimates of the Contribution of Information Technology to Consumer Welfare," MIT Sloan School Working Paper 3647-094, January 1994.
http://www.caldera.com/openstore/index.html, visited 30 August 1998.
"Making DOS Work Overtime," PC Magazine , 16 October 1990, p.196-227.
Central Point Software, PC Tools Shells 5.5 Manual , 1989.
The Computer Language Co. Inc., Computer Desktop Encyclopedia , Computer Select, CD-ROM. July 1998.
Dataquest, Personal Computers Worldwide: Market Statistics , 20 April 1998, Table
"SCO OpenServer Desktop System (Rel.5.0.4),"
1998, Computer Select, CD-ROM, July 1998.
"SCO UnixWare Personal Edition (V.2.1),"
1998, Computer Select, CD-ROM, July 1998.
"The Deal of the Century," History of the Microcomputer Revolution , 1995, http://exo.com/~wts/mits0013.HTM, visited 30 August 1998
"Windows 95: The Verdict is in," PC Magazine August 1995, pp. 86+.
Goodman, John M. and Eva, Elizabeth,
"MS-DOS version 6.22," Infoworld 29 August 1994, pp. 87+.
Information Access Company, Computer Select , CD-ROM, August 1998.
Insight, http://www.insight.com, visited 5 August 1998
International Data Corporation, Client Operating Environments: 1998 Worldwide Markets and Trends , C2348, May 1998, Table
International Data Corporation, Worldwide PC Forecast Update, 1997-2002 , C2346, May 1998, Table
"GeoWorks Ensemble Lays the Foundation for a Brave New OS," PC Magazine, 12 February 1991.
Lexikon Services, The History of Computing, 1998.
"The Gary Kildall Legacy," Amateur Computer Group of New Jersey , 1995, http://www.acgnj.org/newslett/newsx011.html, visited 28 August 1998.
"CP/M: A Family of 8-and 16-bit Operating Systems,"
http://www.maxframe.com/gary & cpm.htm, visited 28 August 1998.
"Informix Goes Linux," Informix Cyber Planet , 1 September 1998, http://www.informix.com/cyb-cgi/Webdriver?Mival=cyberzine_view_Story & ID=1634.
Manes, Stephen and Andrews, Paul, Gates , Touchstone (New York) 1994.
"Market Price and Income Elasticities of New Vehicle Demands," Review of Economics and Statistics , 78(3), August 1996, pages 543-47
"When All You Need is the Fax," Computer Shopper , December 1994, p.
Microsoft Corporation, Microsoft Annual Report, 1997 Fiscal Year , 1997.
Microsoft Corporation, Getting Started: Microsoft Windows 98 , 1998.
Microsoft Corporation, Introducing Microsoft Windows 95 1995, pp. 56-57.
Microsoft Corporation, Microsoft MS-DOS: Users Guide and Reference for the MS-DOS Operating System Version 5.0 , 1991.
Microsoft Corporation, Microsoft Windows Version 3.1 Users Guide: For the Microsoft Windows Operating System , 1992.
Microsoft Corporation, MS-DOS Operating System Version 4.01 Users Guide, Gateway 2000 Edition , 1988.
Microsoft Corporation, Microsoft Windows: Users Guide for the Windows Graphical Environment Version 3.0, Gateway 2000 Edition , 1990.
Microsoft Corporation, MS-DOS Operating System Version 4.01 Users Guide , 1988.
"Concise Users Guide: Microsoft MS-DOS 6.22," Users Guides Microsoft MS-DOS and Microsoft Windows for Workgroups Operating Systems , 1994.
Users Guide: Microsoft Windows, & quote; Users Guides Microsoft MS-DOS and Microsoft Windows for Workgroups Operating Systems , 1994, pp. 15-21.
Miller, Michael J.,
"The Big Picture," PC Magazine , 26 September 1995, p. 108.
Mills, Elinor and Schwartz, Ephraim,
"Word Perfect 8 for Linux Targets Corporate IT Users," Infoworld , 25 May 1998.
"Pegging the Cost of an OS," PC Week , May
4, 1998, p.
"TCP/IP Stack Vendors Reborn," Network World, 17 August 1998, p. 28.
23 July 1998, p. NEW07230018.
"Ericsson, Motorola, Nokia & Psion Create Symbian Alliance,"
24 June 1998, p.
Newton, Harry, Newtons Telecom Dictionary , 1998, Computer Select, CD-ROM, Information Access Company, July 1998.
Parker, P. M. and Neelamegham, R.,
"Price Elasticity Dynamics over the Product Life Cycle: A Study of Consumer Durables,"
INSEAD Working paper 96/71/MKT, Table 3.
PC Connection, http://www.pcconnection.com.
PC Magazine ,
"The History of Windows,"
25 June 1998.
PC Magazine ,
30 October 1990, pp. 65+.
PC Magazine ,
21 April 1998.
PC Webopedia ,
1998, http://webopedia.internet.com/TERM/P/POSIX.html, 1
PC Week ,
"A Linux World?"
31 July 1998.
PC Week ,
Microsoft Pulls Last Plug on Windows
1 September 1997, p. 51.
"Operating in a New Environment," PC Magazine, 25 February 1986, p. 115
Polsson, Ken, Chronology of Events in the History of Microcomputers , 4 August 1998. http://www.islandnet.com/~kpolsson/comphist.htm visited 30 August 1998.
"Red Hat Linux/Intel (Software),"
http://www.redhat.com/products-details.phtml?id=rhl-intel, visited 5 August 1998.
"Lantastic," PC Magazine , 25 May 1993, p. 215.
Rosch, Winn L.,
"The Better Operating System?" PC Magazine, 12 February 1991, p. 245.
"OS/2 2.0: Does It Fulfill the Promise?" PC Magazine , April
28, 1992, p.
"SCO UnixWare Personal Edition,"
http://www.sco.com/unixware21/perserver/, visited 5 August 1998.
Sibley, David S.,
"Declaration of David S. Sibley," United States of America v. Microsoft , 15 May 1998.
Spooner, John G.,
"Jupiter Handhelds Seek out IT Space," PC Week , 4 August 1998, http://www.zdnet.com/pcweek/news/0727/3lejupit.html, 31 July 1998.
"Stata Single-User and Volume-Purchase Pricing,"
(email) 24 August 1998.
Stone, M. David,
"Task Switchers: The Hot Key to Fast Application Access," PC Magazine , 30 October 1990, pp. 175+.
Stone, M. David,
"DESQview," PC Magazine , 25 February 1986, pp. 115-117.
"Sun Solaris Operating Environment Version 2.6,"
http://www.sun.com/software/white-papers/wp-solaris2.6, visited 5 August 1998.
Tay, Richard S. and McCarthy, Patrick S.,
"Demand Oriented Policies for Improving Market Share in the U.S. Automobile Industry," International Journal of Transport Economics 18 (1991), 151-166.
"The THEOS Operating System,"
http://www.theos-software.com/products/t32_os.htm visited 28 August 1998.
"THEOS 32 Version 4,"
http://www.theos-software.com/products/t32_development.htm, 28 August 1998.
Trio Company of Cheektowaga,
"CP/M File & Document Conversions,"
http://www.triousa.com/Convert/CPM/index.html, visited September
United States of America, Department of Justice,
"Complaint," United States of America v. Microsoft , 18 May 1998, p. 1.
Van Arsdale, Steve, and Sisk, Johnson,
"Choosing the Pick of the Litter," PC Magazine , 17 April 1984, p. 321.
"The Be-All-You-Can-Be OS," Byte, December 97, pp. 137-138.
"What is Linux?"
23 July 1998, http://www.zdnet.com/zdnn/stories/zdnn_display/0,3440,2120842,00.html, 19 August 1998.