Looking at Linux: Explaining the Open Source Revolution in Terms of Network Economics
Introduction: A Shift in the Market
A New Paradigm: The Open Source Paradox
The Microsoft Corporation has been the five-hundred-pound gorilla of the software industry for at least the past decade. Microsoft's founder and CEO, Bill Gates, has become legendary for his ruthless business savvy, allowing Microsoft to gain, if not dominate, a major chunk of the market share in most major software markets. This control is nowhere more evident than in the market for operating systems, the bedrock software on which a computer system depends to run all other applications. While Microsoft may have had to struggle in other areas, for most of recent history its Windows OS has been the software almost no modern computer could live without, with a few exceptions such as the struggling Macintosh line. For years, the guaranteed revenue from sales of Windows kept Microsoft's revenue constantly expanding just as fast as the computer industry expanded, allowing Microsoft to take risks in other sectors using the profits from its one unstoppable product (Auletta xx-xxi).
Thus, it was shocking when many major computer hardware manufacturers, including old Microsoft allies, began to neglect Microsoft's Windows NT software, instead producing servers running a nascent OS known as Linux (Auletta 393-94). This defection is only the latest in a series of slights Microsoft has suffered as its once-unstoppable OS monopoly shows signs of eroding under the tide of the "Linux revolution"; many businesses and governments are increasingly shifting their loyalties to Linux. Despite efforts by Microsoft and other software manufacturers to discount Linux's potential as a major OS, a set of internal memoranda called "Halloween Documents" detailed the extreme apprehension among Microsoft's top brass at possible competition from Linux. The documents contain many planned PR strategies designed to discredit Linux as poorly supported, unreliable, and irrelevant to most consumers' needs. This campaign results from fears of Linux's power to severely threaten the Windows monopoly(Raymond 2003).
Linux developers' corporate leadership has not organized a campaign to develop a counterstrategy against Microsoft. Indeed, the set of Linux developers as a whole have no corporate leadership, and, in fact, no corporate ties at all. Conceived and informally organized by its namesake, Linus Torvalds, the Linux project nonetheless continues to exist as a pool of code constantly being added to and modified by hobbyists and small private companies, who release their contributions for free. Various companies, including superstar start-up Red Hat, then select particular versions and patches they find most desirable and release them as a packaged "distribution" of Linux. At no point, however, does anyone gain proprietary ownership of the Linux code, and no royalties are ever charged or paid on any of Torvalds' original code or any of the myriad additions and revisions subsequently made to it. This practice is legally stipulated by the GPL (GNU Public License) under which Linux is copyrighted. Linux has thus become the poster child of what is known as the "free software" or "open source" movement, advocating the utility of a software development model without copyright restrictions or royalties in which anyone may write patches and develop derivative versions(Sullivan 77).
How does this business model work? It seems to contradict basic economic principles; producers of code are not directly compensated for their work and no firm has responsibility for the ownership of the code. Shouldn't the Linux community be suffering from the tragedy of the commons? The answer appears to be no, and the explanation lies in an analysis of a new type of market, the network economy. The open source software movement is driven by an understanding that the traditional model of compensating programmers by selling software as a commodity is misguided. The true value of software ought to be measured not in commodified code but in services provided by developers to users. Under such a model, the market penetration afforded by providing access to software without a per-user charge greatly increases rather than decreases the opportunities for developers' compensation by increasing the number of opportunities for an exchange of services. This model is a far more efficient and effective way to provide incentives to improve software than the commodity model, as Linux's success demonstrates.
A New Economy
Economies of Scale
We must be careful not to confuse the issue by speaking as though Linus Torvalds is in the business of assembling PCs, building skyscrapers or farming wheat. Unlike such physical goods, an operating system is a type of software, and if we consider software as a commidified good, it falls under the traditional category of an information good. An information good suffers from relatively high fixed costs but extremely low variable costs, compared to physically consumed goods; it costs a great deal originally to create a program, but once the program has been created the cost of making another copy of the program is relatively small.
Most information producers thus experience something like a traditional economy of scale, where the average cost of selling software decreases as the total number of copies sold increases. Furthermore, a large portion of fixed cost is sunk; once a company invests the time and energy to write a program, the program can never be unwritten and the resources spent to create it can never be recovered. On the other hand, the marginal cost tends to increase slowly and linearly; the physical media that generate these costs range from easily mass-produced, in the case of CD-ROMs, to requiring no physical production at all, in the case of Internet downloads (which only require a slowly upward-sloping cost to maintain servers and manage network traffic) (Shapiro and Varian 3-7).This cost structure means that there is never much incentive for any firm to exit the market, since the majority of fixed cost will not be recovered; instead, whoever has rights to distribute software will continue doing so as long as revenue covers manufacturing costs.
Economies of Networking
The demand side of an information good demonstrates an even more interesting phenomenon: the so-called economy of networking. The value consumers give to any information good tends to increase as the market share of that information good increases, because the information derives much of its utility from its facilitation of communication and cooperation between consumers. If an OS is more popular, more people will have contact with it, more people will write software with it, and it will be easier to learn how to use it. Coordination between systems is much easier when the OS is the same, and cooperation on projects is much easier when both users have the same software and working environment (Shapiro and Varian 173-75).
Economists characterize such an economy of networking as a "tippy" market, and the software industry is a prime example. Unlike traditional economies of scale, which suffer negative feedback effects as they increase quantity beyond minimum efficient scale of production, economies of networking experience positive feedback as they increase production, because of the previously mentioned low marginal costs combined with a demand that may actually increase with quantity produced. As a corollary, when a software's sales decline below a certain point, the "virtuous cycle" of increasing demand becomes a "vicious cycle" of decreasing demand, as the network of users of a software title reaches the point where it is not large enough to justify users opting into it, and demand drops drastically as users opt for one of the available substitutes.
Thus, a firm that manages to gain control of the market early on will only increase this control until it dominates. Its market dominance is unstable, though, and likely to be overcome by a competing firm that experiences the same meteoric rise in fortunes if it can develop a product that has enough significant advantages over the currently dominant one. This is always possible, if difficult, thanks to the aforementioned strong disincentives for any firm to completely exit an information market (Shapiro and Varian 182-184).
These network effects are reinforced by the complementarity inherent to information goods. Most information goods are not neatly separated into submarkets, but function as systems; if a customer buys the Linux kernel, she will probably buy Linux operating system tools that function well with Linux. The complexity of software means that a certain product depends very much on the quality and compatibility of the products used with it for its utility, so a popular central OS or kernel like Torvalds' Linux kernel will make the subsidiary components of a Linux distribution more popular as well (Shy 2).
Thus, much more so than a traditional economy of scale in industry, information markets experience an extremely sharp economy of scale reinforced from the nature of both the supply and demand side. Indeed, most economists regard the true cost of software production only as the original fixed investment the cost of writing the software and dismiss the cost of distribution as far less relevant to a firm's ability to compete. Once a useful good is produced, network externalities, the additional utility to the whole consumer base attained with each purchase of the product, push demand to increase rapidly (Shy 53).
Commodification and perfect competition
Thus, if we only examine information goods as a commodity bought and sold, once the information has been created and the original fixed costs sunk, there is a powerful incentive for other companies to attempt to enter the market for that particular good, which will have very low cost of production. If this entry is allowed to occur, prices ought to be forced down until all companies selling this good are making zero economic profit, and information markets become a perfectly competitive market similar to agricultural markets; everyone sells the same product at the same price, with no one gaining an advantage over anyone else. This situation does, in fact, happen with many types of information, such as maps, stock prices, or telephone directories (Shapiro and Varian 24).
Software firms generally attempt to avoid allowing the market to become perfectly competitive and to preserve their ability to make monopoly profit by petitioning the government for intellectual property legislation such as copy-protection. This legislation allows them to continue to sell differentiated products, developing and selling a word processor different from and, they claim, better than that of the competitor. Unfortunately, these copy-protection schemes are often costly to implement and present a cost to the consumer by making products more difficult to use (Shy 65-70). Efficiency is lost as each company attempts to gain monopoly power over the others by creating a software product that solves the same problems already solved by competitors' software and by diverting many resources to marketing and bells and whistles to make its product appear unique.
The result is a modern software industry where many duplicate software titles exist for each task. Dominant firms exploit their monopoly advantage and push prices up; network externalities prevent competing firms from engaging in effective price wars with monopolists, so they instead attempt to steal firms' monopoly position through marketing, driving up their own costs. The result is increasing costs to consumers as they face a choice between paying very high prices for dominant software brands (Microsoft and Apple being the best examples, the OS market being the most prone to "tipping") or else choosing between a confusing array of more-or-less equivalent alternatives. There is, however, a way to sidestep this whole problem.
An Open-Source Revolution
Open source software fundamentally operates by decreasing the fixed cost of production to minimum levels, by eliminating the costs necessary to enforce ownership of royalties and instead compensating programmers as providers of secondary services in a commodity market, maximizing network externalities. Rather than pitting different support networks centered around different software titles against each other, programmers are free to work together to build the most interconnected support network around the best available software, competing against each other to provide the best service in doing so.
The reason for inefficiency in the current market leading to slower development and higher consumer costs than necessary is a game-theoretical one; competing firms will engage in actions harmful to the overall efficiency of the industry by reducing network externalities in order to increase their total payoff (Shy 53). The most obvious example is the positive network externality generated by standardized software; if all firms agree to produce software according to a known standard, software becomes much more valuable to the consumer, as the opportunity cost of using different accessory software becomes lower with increases in compatibility. Proprietary firms have a disincentive to standardize their products, however, as the less unique their software is the harder it is for them to retain their share of the market (Shapiro and Varian 223-229).
This behavior, endemic in the proprietary software industry, becomes irrelevant to an open source software developer. Open source software developers do not charge royalties for their product, giving them no incentive to try to shut out others' work from the market, and thus open source software is characterized by a high reliance on others' work (Raymond 2003).
In general, duplication of labor is thus avoided in open source software; the necessity to keep trade secrets is removed, and developers are able to build directly upon each other's work. As we saw before, this collaboration means that the software has an extremely high chance of outcompeting any existing software that has higher fixed development costs, and at this point Linux is being developed, modified and maintained at a much faster rate than Microsoft Windows (Raymond 2000).
Selling Something Else
The key to understanding the incentive that open-source developers have to work on products "for free" is to understand the nature of a network economy, specifically an information economy. The ease of creating information products means consumers suffer an overabundance of choice, and thus the most precious commodity is one's time spent searching for a software solution to a problem (Shapiro and Varian 24-27). Thus, there is a strong incentive for educated users to try to create patches themselves in order to solve a problem they face during their use of the software, and given the broad labor base of the imaginary "firm" producing Linux, all that is needed from each individual is a vanishingly small contribution to the total finished code. The amount of money that Microsoft programmers demand as compensation is largely a result of the cloistered, concentrated level of work required of programmers in a corporate environment, where each member of a project team has a much higher level of responsibility than would be required of them in an open source project (Benkler 367).
Looking at the programming industry reveals that most of the work programmers do can be classified more neatly under the banner of "maintaining" than "producing" software. Programmers look for bugs, analyze performance and teach the use of existing software, services for which they are paid without being given royalty control of their work. These services make up roughly 95 percent of all programming labor as opposed to the 5 percent that goes into a copyrightable product. Linux's development model reduces all such design jobs into routine maintenance jobs, taking advantage of the power of software to be built bit-by-bit. Linux programmers find that, the more Linux distributions exist, the more they are paid by companies to solve some particular problem with Linux or to adapt Linux to their company's specific situation, providing a strong networking incentive to be an active part of Linux's development community (Benkler 367).
It is by harnessing these network externalities that companies can stand to profit while serving customers best. The commodity in a market for services rather than for proprietary software is human creativity rather than packaged bytes. Such a market is likely to be far more competitive, as individual human consultants on software are a more rivalrous good than a single software package; there is a higher incentive to enter such a market, and we already see companies like Red Hat doing so and making a profit by selling Linux distributions and guaranteeing support for them, despite Linux's relatively low market share. Ironically, because of the original conception of software being a buy-once commodity like an automobile or a refrigerator, software companies offered the same free technical support services as auto dealers. This free service makes technical support suffer from a tragedy of the commons effect; tech support becomes a major expense for such companies, which get no direct compensation for it, and frustrated consumers find that there aren't enough personnel to attend to their needs.
The situation is analogous to, say, the stock market, where the mechanisms of the market itself and the various options available to the investor are beyond the expertise of the average layperson. In such situations, businesses are willing to pay quite high prices for the time of educated professionals who provide the overall service of creating and managing a portfolio. The growing industry of providing overall software "solutions" by picking through the available software options and providing support and management for them could become far more efficient if the process of creating and modifying code were integrated with higher-level services rather than separated into an artificial commodity market as it is now. It is the rivalrous, excludable service of creating, modifying, adapting and using software code to provide specific services that businesses ought to pay for, and not the nonrivalrous, artificially excluded good of software code itself (Benkler 367).
Auletta, Ken. World War 3.0: Microsoft and Its Enemies. New York: Random House, 2001.
Benkler, Yochai. "Coase's Penguin, or, Linux and the Nature of the Firm." Yale Law Journal, (December, 2002) p. 367-378.
Raymond, Eric S. "The Halloween Documents." 2003. Accessed March 28, 2003. http://www.opensource.org/halloween/
Raymond, Eric S. "The Magic Cauldron." 2000. Accessed March 28, 2003. http://www.catb.org/ ~esr/writings/cathedral-bazaar/magic-cauldron/
Shapiro, Carl, and H. R. Varian. Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press, 1999.
Shy, Oz. The Economics of Network Industries. Cambridge: Cambridge University Press, 2002.
Arthur Chu '06 is an honors history major and an honors economics and Chinese minor. He chose to explore the "Linux phenomenon" because he is "fascinated with the ways the digital revolution has changed the way people live their lives, do business and create." Arthur states, "I continue to be interested in the open-source movement and the implications it has for the future of capitalism." About the writing process, he says, "Writing for me is a fairly holistic process; I write quickly and all at once, preferring to set aside a block of time to write out, reread, revise, and tinker rather than methodically planning out the writing process. The most difficult part of writing is getting the motivation to start, which usually involves bouncing ideas off of friends until I get a spark of inspiration." This paper was written for Professor Leah Smith's Economics 1: Introduction to Economics.