Network Externalities (Effects)

S. J. Liebowitz

Stephen E. Margolis

S. J. Liebowitz: Management School, University of Texas at Dallas, Richardson, Texas 75083

Stephen E. Margolis: Department of Economics, North Carolina State University, Raleigh, North Carolina 27695

1. What are Network Effects?

Network externality has been defined as a change in the benefit, or surplus, that an agent derives from a good when the number of other agents consuming the same kind of good changes. As fax machines increase in popularity, for example, your fax machine becomes increasingly valuable since you will have greater use for it. This allows, in principle, the value received by consumers to be separated into two distinct parts. One component, which in our writings we have labeled the autarky value, is the value generated by the product even if there are no other users. The second component, which we have called synchronization value, is the additional value derived from being able to interact with other users of the product, and it is this latter value that is the essence of network effects.

An illustration: As this article was being written, commentators are speculating on whether Apple computer will survive, since its network (base of users) is shrinking, some think, below a minimum acceptable level. Since the actual quantity of Apple computers sold is still among the very largest of personal computer manufacturers which should allow Apple to take advantage of economies of scale in production, and the computers are not thought to be deficient in terms of quality, any lack of viability must be due to the fact that the network of Apple computer users is small. In other words, the synchronization value of Apple computers is thought to be too low.

First a definitional concern: Network effects should not properly be called network externalities unless the participants in the market fail to internalize these effects. After all, it would not be useful to have the term ‘externality’ mean something different in this literature than it does in the rest of economics. Unfortunately, the term externality has been used somewhat carelessly in this literature. Although the individual consumers of a product are not likely to internalize the effect of their joining a network on other members of a network, the owner of a network may very well internalize such effects. When the owner of a network (or technology) is able to internalize such network effects, they are no longer externalities. This distinction, first discussed in Liebowitz and Margolis (1994) now seems to be adopted by some authors (e.g. Katz and Shapiro 1994) but has not been universally adopted.

Putting aside definitional concerns, the import of network effects comes largely from the belief that they are endemic to new, high-tech industries, and that such industries experience problems that are different in character from the problems that have, for more ordinary commodities, been solved by markets (Katz and Shapiro 1985, Farrell and Saloner 1985, Arthur 1996). The purported problems due to network effects are several, but the most arresting is a claim that markets may adopt an inferior product or network in the place of some superior alternative. Thus if network effects are a typical characteristic of modern technologies, the theory suggests that markets may be inadequate for managing the fruits of such technologies.

The concept of network externality has been applied in the literature of standards, where a primary concern is the choice of a correct standard (Farrell and Saloner 1985, Katz and Shapiro 1985, Besen and Farrell 1994, Liebowitz and Margolis 1996). The concept has also played a role in the literature of path dependence (Arthur 1989, 1990, David 1985, Liebowitz and Margolis 1990, 1995a.).

The literature has identified two types of network effects. Direct network effects have been defined as those generated through a direct physical effect of the number of purchasers on the value of a product (e.g. fax machines). Indirect network effects are "market mediated effects" such as cases where complementary goods (e.g. toner cartridges) are more readily available or lower in price as the number of users of a good (printers) increases. In early writing, however, this distinction was not carried into models of network effects. Once network effects were embodied in payoff functions, any distinction between direct and indirect effects was ignored in developing models and drawing conclusions. However, our 1994 paper demonstrates that the two types of effects will typically have different economic implications. It is now generally agreed (Katz and Shapiro, 1994) that the consequences of internalizing direct and indirect network effects are quite different. Indirect network effects generally are pecuniary in nature and therefore should not be internalized. Pecuniary externalities do not impose deadweight losses if left uninternalized, whereas they do impose (monopoly or monopsony) losses if internalized. An interesting aspect of the network externalities literature is that it seemed to ignore, and thus repeat, earlier mistakes regarding pecuniary externalities. (For the resolution of pecuniary externalities see Young (1913), Knight (1924), and Ellis and Fellner (1943).)

Concern about marginal adjustment of the level of network activity has not been the primary focus of the network externality literature. Instead, this literature has focused primarily on selection among competing networks. Our discussion below follows this relative emphasis. We briefly consider the issue of levels of network activities in section two, then turn to the choice of networks.

2. Levels of Network Related Activities

Harvey Liebenstein’s work on bandwagon and snob effects (1950) anticipates much of the argument regarding network effects. His main result is that demand curves are more elastic when consumers derive positive value from increases in the size of the market.

One branch of the modern network literature would easily fit in the Liebenstein framework. Such research has reexamined various economic models with network effects introduced. For example, an analysis of the impacts of unauthorized software copying will change when network effects are introduced. Copying increases the size of a network, increasing the value to authorized users, so that any harm from unauthorized copying will likely be mitigated.

The difference between a network effect and a network externality lies in whether the impact of an additional user on other users is somehow internalized. Since the synchronization effect is almost always assumed to be positive in this literature, the social value from another network user will always be greater than the private value. If network effects are not internalized, the equilibrium network size may be smaller than is efficient. For example, if the network of telephone users were not owned, it would likely be smaller than optimal since no agent would capture the benefits that an additional member of the network would impose on other members. (Alternatively, if the network effects were negative, a congestion externality might mean that networks tend to be larger than optimal.) Where networks are owned, this effect is internalized and under certain conditions the profit maximizing network size will also be socially optimal. (For which see our paper, 1995b.)

Perhaps surprisingly, the problem of internalizing the network externality is largely unrelated to the problem of choice between competing networks that is taken up in the next section. In the case of positive network effects, all networks are too small. Therefore, it is not the relative market shares of two competing formats, but rather, the overall level of network activity that will be affected by this difference between private and social values. This is completely compatible with the literature on conventional externalities. For reasons that we will expand on below, this is a far more likely consequence of uninternalized network effects than the more exotic cases of incorrect choices of networks, standards or technologies.

Network size is a real and significant issue that is raised by network effects. Nevertheless, this issue has received fairly little attention in the contemporary literature of network externality, perhaps because it is well handled by more conventional economic models.

3. Choice Among Competing Networks Under Increasing Returns.

The literature on network externalities challenges economists' traditional use of decreasing returns and grants primacy to economies to scale. Positive network effects, which raise the value received by consumers as markets get larger, have impacts that are very similar to conventional firm-level economies of scale. If we start an analysis with the assumption that firms produce similar but incompatible products (networks), and that the network effects operate only within the class of compatible products, then competitors (networks) with larger market shares will have an advantage over smaller competitors, ceteris paribus. If larger competitors have a forever widening advantage over smaller firms, we have entered the realm of natural monopoly, which is exactly where most models that address network and standards choices find themselves.

It is critical to note, however, that network effects are not in general sufficient for natural-monopoly-type results. In cases where average production costs are falling, constant, or nonexistent, network effects would be sufficient for a result of natural monopoly. Many, if not most, models in this literature ignore production costs and thus with any assumption of positive network effects are unavoidably constructed as instances of natural monopoly. But notice that if production costs exhibit decreasing returns, and if these decreasing returns overwhelm the network effects, then natural monopoly is not implied, and competing incompatible networks (standards) will be possible.

Though economists have long accepted the possibility of increasing returns, they have generally judged that except in fairly rare instances, the economy operates in a range of decreasing returns. Some proponents of network externalities models predict that as newer technologies take over a larger share of the economy, the share of the economy described by increasing returns will increase. Brian Arthur has emphasized these points to a general audience: "[R]oughly speaking, diminishing returns hold sway in the traditional part of the economy – the processing industries. Increasing returns reign in the newer part – the knowledge-based industries … They call for different management techniques, strategies, and codes of government regulation. They call for different understandings" (Arthur 1996: 101)

If the choice of a standard or network is dominated by natural monopoly elements, then only one standard will survive in the market. It thus is of great importance that the standard that comes to dominate the market also be the best of the alternative standards available. Traditionally it has been assumed that the natural monopolist who comes to dominate a market will be at least as efficient as any other producer, but this assumption is specifically questioned in the newer literature. Specifics differ across the many versions of this problem that appear in the literature. The recurring issue, however, is that we lose the usual assurances that the products that prevail in markets are the ones that yield the greatest surpluses.

The mere existence of network effects and increasing returns is not sufficient to lead to the choice of an inferior technology. For that, some additional assumptions are needed. One common assumption that can generate a prediction of inefficient network choice is that the network effect differs across the alternative networks. In particular, it is sometimes assumed that the network type that offers the greatest surplus when network participation is large is the one that offers the smallest surplus when participation is small. This condition, however, is not likely to be satisfied, since synchronization effects are likely to be uniform. For example, if there is value in a cellular telephone network becoming larger, this should be equally true whether the network is digital or analog. Similarly, the network value of an additional user of a particular videorecorder format is purported to be the benefits accrued by having more opportiunities to exchange video tapes. But this extra value does not depend on the particular format of videorecorder chosen. If network effects are the same for all versions of a given product, it is very unlikely that the wrong format would be chosen if both are available at the same time. (See also "Path Dependence" in this dictionary for a further discussion of this issue.)

4. Common Restrictions in Network Effects Models

As we have noted, network externality models often feature particular outcomes: Survival of only one network or standard, unreliability of market selection, and the entrenchment of incumbents. In formal models, these results follow inevitably from assumptions that are common simplifications in economic theory and that appear to be relatively unrestrictive. As applied in these network models, however, these assumptions are both critically responsible for the results and unappealingly restrictive.

Two important limitations of many network externalities models are the assumptions of constant marginal production cost and network value functions that rise without limit [see, for example, Chou and Shy (1990: 260), Church and Gandal (1993: 246), Katz and Shapiro (1986: 829), and Farrell and Saloner (1992: 16)]. Matutes and Regibeau (1992) consider issues of duopoly and compatibility under a similar structure. Such assumptions install an inexhaustible economy of large-scale operation. If a network size can reach a point where additional participation does not provide additional value to participants, increases in scale would no longer be advantageous and it would then be possible for multiple networks to compete on an even basis.

Without investigation, it is unreasonable to accept that the law of diminishing marginal product is somehow suspended in new-technology industries. While the scale properties of a technology pertain to the simultaneous expansion of all inputs, it seems evident that resource limitations do ultimately restrain firm size. Economists have long supposed that limitations of management play a role in this, a relationship formalized in Radner (1992). The capacity constraints evident in the crisis at America On-Line in early 1997 exemplifies the influence of resource limitations on the performance of a network.

Another limitation is the common assumption that consumers are identical in their valuations of competing networks. Once heterogeneous tastes are allowed it becomes feasible for competing networks to coexist with one another even though each exhibits natural monopoly characteristics (e.g. If some computer owners much prefer Macintoshes and others much prefer the PC, they could both coexist).

A further restriction is the symmetric value received by consumers when another consumer joins a network. If economists, for example, much prefer to have other economists join their network as opposed to, say, sociologists, then it would be possible for economists to form a coalition that switches to a new standard even if the new standard failed to attract many sociologists. This latter point will prove to be of great importance when examining empirical examples of choosing the wrong standard, where large entities such as multinational firms and governments play an important role.

6. The Empirical Relevance of Network Effects and Increasing Returns

The confluence of network effects, increasing returns, and market outcomes may be spurious. Although many technologies have tended to evolve into single formats (e.g. home-use VCRs are almost all of the VHS variety) some portion of these may actually have evolved for reasons having little to do with either network effects or increasing returns. We should not be surprised to find that where there are differences in the performance of various standards, one may prevail over the others simply because it is better suited to the market.

First of all, the extent (and symmetry) of network effects may be much more limited than is commonly assumed. For example, in the case of spreadsheets and word processors, it may be quite important for a small group of collaborators to use identical software so as to be perfectly compatible with each other. Similarly, compatibility may be important for employees within a firm. But compatibility with the rest of the world may be relatively unimportant, unimportant enough to be overwhelmed by differences in preferences, so that multiple networks could survive. Networks that serve niche markets well (such as wordprocessors specializing in mathematical notation), might not be significantly disadvantaged by network effects.

As an illustration, consider the empirical examples of tax software and financial software. By far the dominant firm is Intuit, with its Turbo-Tax products and Quicken financial software. This market seems to have tilted strongly toward a single producer. Yet network effects should be virtually nonexistent for these products. Consumers do not exchange this type of information with one another. A better explanation that is consistent with the product reviews in computer magazines is that these products are just better than the alternatives.

Similarly, for large firms, compatibility within the firm should be of great importance, but compatibility outside of the firm should be of little consequence. For many products where the majority of customers are large firms, producers will not encounter natural monopoly elements since there may be little or no network advantage in selling to multiple firms.

Regarding increasing returns in production, it is true that the past decades have evidenced a number of technologies that have experienced enormous declines in prices and tremendous growth in sales. Nevertheless, it is not clear that this is the result of increasing returns. Since bigger has been cheaper, it has often been assumed that bigger causes cheaper. But an available alternative explanation is that as technologies have advanced with time, the average cost curves are shifting down. If that is the case, the implied causality is reversed: Cheaper causes bigger. Consider for example, the history of old technologies, such as refrigerators and automobiles. These industries, currently thought to exhibit conventional decreasing returns, experienced tremendous cost decreases, along with tremendous increases in utilization, early in their histories. (For a discussion of an earlier episode of this confusion, see Stigler 1941).

7. Policy Implications

The theory of network externality is currently playing a role in several antitrust actions, the most prominent of which are the investigations of Microsoft by the Justice Department over various aspects of Microsoft’s behavior including its attempted purchase of Intuit, the inclusion of the Microsoft Network as an icon in Windows 95, Microsoft’s attempt to wrest control of the web browser market from Netscape and so forth. The claim seems to be that since markets cannot be relied upon to choose the best standards or bring about the right networks, governments might wish to investigate and control firms’ efforts to make standards or establish networks. Several of the theorists involved in this literature have consulted for parties involved with these investigations. For example, Brian Arthur and Garth Saloner were two of the authors in the amicus curiae brief against Microsoft presented by four anonymous parties (thought to be four well-known and generally well-heeled competitors of Microsoft).

Clearly the potential to misuse such antitrust theories by competitors unable to win in the marketplace is very great, not unlike various theories of predation. Since the empirical support for this theory is so weak, it appears at best to be premature and at worst simply wrong to use this theory as the basis for antitrust decisions.

In a related vein, for those cases where network effects are important, the role of copyright and patent law may be cast in a new light. First, networks are likely to be too small if network effects are not internalized. Intellectual property laws are one means by which such network effects can be internalized, since ownership is an ideal method of internalization.

The possibility that networks can compete with each other suggests a further consideration regarding intellectual property law. Where one standard is owned and another is not, we can have less confidence that an unowned but superior standard will be able to prevail against an owned standard, since the owner of a standard can appropriate the benefits of internalizing any network effects. Although we do not have any evidence that inferior standards have prevailed against superior standards, this may be in large part because most standards are supported by companies who have some form of ownership such as patent, copyright, or business positioning. The greatest chance for some form of third-degree path dependence (see Path Dependence, this volume) to arise would be if an unowned standard with dispersed adherents were to engage in competition with a standard that had well defined ownership. Further research in this area is needed before any firm conclusions can be drawn, however.

8. Conclusions

Network effects are undoubtedly real and important phenomena. The popular and very compelling example is the telephone network. Who would deny that the value of phone service depends heavily on the number of other people who have phone service. Contemporary technologies expand that example enormously.

The enthusiasm for recognizing and understanding these phenomena should not, however, lead us to inappropriate or premature conclusions. As we have noted above, there are distinctions and reservations that ought to be maintained. The first and broadest is that between network effects and network externalities. A further distinction is between pecuniary externalities and real ones. Even for the set a real externalities, it is important to note the distinction between the problem of network size and that of network choice, the boundedness of the network effect, the likely symmetry of network effects for alternative products, the ability of large consumers to self-internalize network effects, and differences in tastes.

Finally, we would urge some reservation about the empirical validity of economies of production scale for many high tech products. If these products have diseconomies of scale at some production level, these production costs may overturn other natural monopoly elements. Improvements in production costs, as with many other economic results, may have more to do with being smarter than with being bigger.







Arthur, W. B. 1989. Competing technologies, increasing returns, and lock-in by historical events. Economic Journal 99: 116-131.

Arthur, W. B. 1990. Positive feedbacks in the economy. Scientific American 262: 92-99.

Arthur, W. B. 1996. Increasing returns and the new world of business. Harvard Business Review: 100-109.

Besen, S. M. and Farrel, J. 1994. Choosing how to compete: Strategies and tactics in standardization. Journal of Economic Perspectives 8: 117-31.

Chou, D. and Shy, O. 1990. Network effects without network externalities. International Journal of Industrial Organization 8: 259-270.

Church, J and Gandal, N. 1993. Complementary network externalities and technological adoption. International Journal of Industrial Organization 11: 239-260.

David, P. A. 1985. Clio and the economics of QWERTY. American Economic Review 75: 332-37.

Ellis H. S. and Fellner, W. 1943. External economies and diseconomies. American Economic Review 33: 493-511.

Farrell J. and Saloner, G. 1985 Standardization, compatibility, and innovation Rand Journal 16: 70-83.

Farrell J. and Saloner, G. 1992. Converters, compatibility, and control of interfaces Journal of Industrial Economics 40: 9-36.

Katz M. L., and Shapiro C. 1985. Network externalities, competition, and compatibility. American Economic Review 75: 424-440.

Katz, M. L., and Shapiro, C. 1986. Technology adoption in the presence of network externalities. Journal of Political Economy 94: 822-841.

Katz, M. L. and Shapiro, C. 1994. Systems competition and network effects. Journal of Economic Perspectives 8: 93-115.

Knight, F. H. 1924 Some fallacies in the interpretation of social cost. Quarterly Journal of Economics 38: 582-606.

Leibenstein, H. 1950. Bandwagon, snob, and Veblen effects in the theory of consumer’s demand. Quarterly Journal of Economics 64: 183-207.

Liebowitz, S. J. and Margolis, S. E. 1990. The fable of the keys. Journal of Law and Economics 33: 1-26.

Liebowitz, S. J. and Margolis, S. E. 1995a. Path dependence, lock-in and history. Journal of Law, Economics and Organization 11: 205-226.

Liebowitz, S. J. and Margolis, S. E. 1995b. Are network externalities a new source of market failure? Research In Law And Economics 17: 1-22.

Liebowitz, S. J. and Margolis, S. E. 1996. Market processes and the selection of standards. Harvard Journal of Law and Technology 9: 283-318.

Matutes, C. and Regibeau, P. 1992. Compatibility and bundling of complementary goods in a duopoly. Journal of Industrial Economics 40: 37-54.

Radner, R. 1992. Hierarchy: The economics of managing. Journal of Economic Literature 30: 1382-1415.

Stigler, G. J. 1941. Production and Distribution Theories, New York: Macmillan Company.

Young, A. A. 1913. Pigou's Wealth and Welfare. Quarterly Journal of Economics 27: 672-86.