Yi-Nung Yang**

May, 1997; revised August 1997

Department of Economics

UTAH STATE UNIVERSITY

Logan, Utah 84322-3530

__________________________________________________

*This paper is the first chapter of Yi-Nung Yang's dissertation.

**Yi-Nung Yang is assistant professor, Department of International Trade, Chung Yuan Christian University, Chung-Li, Tao-Yuan, Taiwan, Republic of China. E-mail: yinyang@ms17.hinet.net.

Some goods and services generate more value when more users consume the same goods and services. Many products are required to be used with other products at the same time. They have little or even no value if they are used in isolation. The consumers using these products constitute networks in which the utility derived from consumption of these goods or services increases as additional consumers purchase the same goods and services. A market characterized by such properties is called a network market in which there exist positive consumption externalities termed network externalities.

For instance, facsimile machines and their communication protocols jointly provide fax services. People who own fax machines form a typical example of a network to exchange information. Computers furnish services because hardware and software are working together. Telephones provide a convenient way for people to communicate with one another as more join the telephone use networks. One should note that some products that are not so-called high technology products may also exhibit features of network products. For examples, camera bodies and lenses provide photographic services (Michael Katz and Carl Shapiro, 1994). Products that are strongly complementary also exhibit positive network effects.

Network products have become more prominent in our modern life as technologies have made progress during recent decades. Many network markets differ from those in traditional industries in their conduct and performance due to the presence of network externalities. The presence of network externalities has proven to be problematic in traditional economic theory (Michael Katz and Carl Shapiro, 1985, 1994) as manifested by extensive research on the subject. However, little formal modeling of firm response to network effects and standardization in the presence of these effects has been completed to date. Therefore, the objectives of this study are to (1) formally analyze the response of a monopolist firm to the evolvement and presence of network effects; (2) determine if there are commitment mechanisms that a monopolist firm can use to control supply of a durable good and modify the downward pressure on price of such goods over time; (3) investigate the role of government intervention to improve network market outcomes; and (4) formally investigate the private and social incentives to achieve international compatibility for a product characterized by network effects.

Network effects or network externalities arise in such markets. A network externality is defined as the increasing utility that a user derives from consumption of a product as the number of other users who consume the same product increases (Michael Katz and Carl Shapiro, 1985). In a broader sense, a network externality is the increase in the net value of an action that occurs as the number of agents taking equivalent actions increases (S. J. Liebowitz and Stephen Margolis, 1995). Examples of products exhibiting network externalities include fax machines, telephone networks, broadcast industry services, credit card networks, and computer hardware and software, etc. Although this positive adoption externality receives most attention in the network literature, negative network externalities, or congestion externalities can possibly arise as network effects.

Indirect network effects arise when the value of a product increases as the number of, or the variety of, the complementary goods or services increases. Computer services exhibit indirect network effects. This indirect network effect in the computer industry is sometimes referred to as the hardware-software paradigm. One may observe that the success of an operating system for personal computers depends not merely on its technological advantages, but also (and often this is more crucial) on varieties of software applications available in the market. Value in use may depend more critically on software applications.

A. Expectations

When consumers choose products in network markets, their expectations play a crucial role on sales of the products or their network components since each consumer's utility depends on the number of others purchasing the same products. For example, those who bought the Betamax standard for videocassette recorders format might have rejected the Betamax system if they had known in advance that only the VHS system would eventually survive.

Rival firms in the network industry thus would try to influence consumers' expectations in order to maximize their profits, especially when consumers have only imperfect information about the size of the installed base in the market. Sales figures are often exaggerated to impress consumers about leadership in the installed base over rivals. The competition between OS/2 and Microsoft Windows in 1992 illustrates a prominent example of firms' intentions of influencing consumers' expectations. IBM and Microsoft both announced wider adoption of their operating systems and disputed the estimated numbers of sales by each other (Stanley Bensen and Joseph Farrell, 1994).

B. Coordination

If coordination between consumers is feasible, it could reduce the risk of choosing the wrong network and enable consumers to enjoy higher network benefits in a larger, sustainable network. However, the difficulty of facilitating coordination among a large group of consumers might impede consonance of consumers' decisions because of the high transaction costs or heterogeneity of preferences.

The incentives for achieving coordination among producers are not always in the same direction. Firms' agreement on a specific standard for their products may attract more users into the network market because users may expect a larger network that is sponsored by a group of firms. Nevertheless, an industrywide standard could possibly intensify the competition among firms since the standard can make products from different suppliers more homogeneous and thus decrease industry profits.

C. Compatibility

Definitions of compatibility vary in the literature (for example, Nicholas Economides, 1989, 1996; Joseph Bailey et al., 1995). In general, it can be said that two products are compatible when the cost of combining them to generate services is free.

If agents cannot coordinate in advance, incompatibility may arise. However, it is misleading to view incompatibility as simply another coordination failure (Michael Katz and Carl Shapiro, 1994). Consumers do not always benefit from compatibility because compatibility can restrict the variety of goods and perhaps constrain future innovations (Carmen Matutes and Pierre Regibeau, 1988; Nicholas Economides, 1989; Richard Gilbert, 1992). When consumers are heterogeneous, they might treat variety and compatibility as substitutes (Joseph Farrell and Garth Saloner, 1986b; Oz Shy, 1996). A consensus of consumers' opinions about compatibility therefore may not arise even if they coordinate with each other prior to their purchase decisions.

Furthermore, the market mechanism may give producers incentives to design their products to be compatible since a compatible design adds value for consumers under some circumstances. But incompatibility is also often observed in markets. The reason is simply because network externalities cannot be internalized through compatibility. An increase in one firm's sales expands the whole network and concurrently raises demand for rival firm's product if their products are compatible. As a result, a firm with a good reputation may oppose compatibility whereas others may favor compatibility (Michael Katz and Carl Shapiro, 1985).

D. Switching Costs

Consumers as well as firms have to face some barriers if they are willing to transfer from one network to another when there is inadequate standardization. Switching costs provide barriers that prevent agents from entering into another network. Switching costs for consumers and firms fall into two categories, social and private costs of switching.

When an agent considers a switch, the agent has to calculate two sorts of costs. One is the private cost due to sunk investments in the original technology. The other is the net network effects, termed the social cost of switching, in which the current network effects the agent is enjoying are compared with the potential network effects expected from switching

For instance, when a MS-DOS user considers switching to another operating systems such as Windows or OS/2, the availability and variety of software applications on the new operating system, and an estimate of how much effort it will take to convert those documents, worksheets, and database from MS-DOS to, say OS/2, must be considered. Producers face a similar situation as consumers. It follows that switching costs impose asymmetric prices between users with and without any sunk investment in an existing technology.

Switching costs may lead to inefficiency by preventing users from adopting a new emerging, superior technology, and the inefficiency is increased if network externalities are significant.

A. Start-Up and Critical Mass

Since the demand for a network good has long been determined to be a function of both its price and the expected size of the network or its complementary network (Michael Katz and Carl Shapiro, 1985, 1994; Stanley Bensen and Joseph Farrell, 1994; Nicholas Economides, 1996), the problem of startup arises when consumers expect that no one would buy the good or that no complementary good would be available in the market. This problem intensifies when the network good has little use value in isolation. Typical examples come from the telecommunications industry such as use of cellular phones. Of course, this is an insignificant problem if a network product has a relatively high intrinsic or stand-alone value.

An interesting concept termed critical mass has been developed in studies of the communications market (Jeffrey Rohlfs, 1974; Shmuel Oren and Stephen Smith, 1981). Existence of critical mass in markets with network effects suggests that a sustainable growth of the network requires a minimal nonzero equilibrium size (Nicholas Economides, 1996). Nicholas Economides and Charles Himmelberg (1995) interpret critical mass as the "chicken and the egg" paradox from observations that the expected network size is too small to induce consumers into the network. In turn, because no consumers are willing to join the network, the expected size of the network is small. They model a multiplicative formulation in network demand and prove that market structure does not affect the existence or the size of minimum feasible networks.

B. Instability of Markets

Stanley Bensen and Joseph Farrell (1994) describe network markets as "tippy," i.e., the markets are prone to instability. One may observe an unstable coexistence of incompatible products such as the classic case in the videocassette recorder market. Dominance of one technology today does not guarantee continued success forever since such instability can happen. Stanley Bensen and Joseph Farrell illustrate a case for such rapid tippiness of competing technologies in the video encryption of cable television programs.

Several reasons explain such instability of network markets. Multiple equilibria, often multiple corner equilibria, in network markets can cause this tippiness (Michael Katz and Carl Shapiro, 1985). Second, instability arises if expected sales rather than current sales matter since today's success cannot ensure tomorrow's. Finally, a disequilibrium network in which the network size does not go beyond the point of the critical mass can be observed.

C. Path Dependence

Problems caused by path dependence in a positive feedback model were raised by Paul David (1985) and W. Brian Arthur (1989). Path dependence induces an inefficiency arising from small differences in initial conditions which lead to outcomes that are likely to be costly to change (S. J. Liebowitz and Stephen Margolis, 1995). Effects of decisions by earlier adopters on the decisions of later adopters are often significant in network markets. The case of the QWERTY keyboard (Paul David, 1985) is a well-known example of the path-dependence problem. The current dominance of the QWERTY keyboard today is not thought to be due to its superiority for typing but because it was invented earlier than the Dvorak keyboard. (Paul David, 1985; Stanley Bensen and Joseph Farrell, 1994; Michael Katz and Carl Shapiro, 1994).

D. Winner with Inferior Technology

Users who have bought a network product that turns out to be an inferior technology may suffer because switching to the superior technology may be costly. The rationale goes as follows. First, high private cost of switching keeps consumers from adopting the superior technology. Second, the existing installed base of current technology makes such a switch difficult due to network effects. Joseph Farrell and Garth Saloner (1985, 1986a) call this "excess inertia," i.e., a Pareto-superior new technology might not be adopted. Inferior technologies can win when the adoption process in technologies is path-dependent. Furthermore, Joseph Farrell and Garth Saloner (1985, 1986a) and Michael Katz and Carl Shapiro (1986a, 1992) formally show that an inferior product may defeat a superior one when consumers expect the superior technology will not be accepted.

A. Adoption Externalities

Positive consumption or demand externality from a communications service was initially described by Jeffrey Rohlfs (1974) and Shmuel Oren and Stephen Smith (1981). They found a subscriber's utility increases as others join the system. Jeffrey Rohlfs (1974) also shows that there are multiple equilibria at any given prices. An inverted U-type of demand function is demonstrated and the startup and critical mass problems are noted by Nicholas Economides and Charles Himmelberg (1995). Michael Katz and Carl Shapiro (1985, 1986a) formally describe this phenomenon as a form of network externality.

The marginal private benefit from joining a network is less than the social benefit when the network market exhibits positive consumption externalities. It follows that the network size in equilibrium may be smaller than the social optimal size. A perfectly competitive market equilibrium will be inefficient due to these externalities. Therefore, traditional welfare analysis may not apply under these conditions. In a game theoretic analysis, Philip Dybvig and Chester Spatt (1983) discuss the welfare implications associated with adoption externalities and suggest potential instruments, such as a subsidy, to remedy this type of externality.

Inefficiencies resulting from adoption externalities arise in other ways. For example, public opinion regarding the network industries suggests that markets tend to be stranded in an inferior technology, i.e., excess inertia exists. Joseph Farrell and Garth Saloner (1985) show that the occurrence of the inertia depends on whether or not there is complete information. This socially excessive resistance to a new superior technology creates inefficiency in the dynamic process of adoption. They further show that the network effect may inhibit innovation (Joseph Farrell and Garth Saloner, 1986a).

Excessive inertia, however, is still debatable on a theoretical basis. For instance, Michael Katz and Carl Shapiro (1992) prove that there may be a tendency to be biased in favor of a new, incompatible technology. This expectation is termed insufficient friction as one can observe that many new, and incompatible technologies, are indeed successfully introduced. They argue that today's users who decide to buy a new but incompatible technology ignore the externalities they impose on yesterday's users who are stranded with the existing technology.

There are several key factors in determining the excessive or insufficient adoption process. Sponsorship of an emerging technology is important because a sponsored technology is more likely to be successful. Consider one new technology with sponsorship and one old technology competitively supplied in which the latter may be unsponsored due to expiration of some patents. A sponsored technology may represent further improvement and lower cost in the future while an unsponsored one seems to have no future. These asymmetries in sponsorship between old and new technologies can determine the outcome of excess inertia or insufficient friction (Michael Katz and Carl Shapiro, 1992, 1994).

The second factor arises in the asymmetrical switching costs between old and new users but many models fail to take this switching into consideration. Given the incompatibility of the old and new technologies, old technology users actually face a higher price than new users. If significant network effects exist, old users resist transferring to a new technology. New users know the expected network would be smaller given this resistance and excess inertia therefore occurs.

B. Intersystem and Intrasystem Competition

Systems competition has attracted considerable attention in the literature. Two categories, intersystem and intrasystem competition, are discussed below.

Intersystem competition characterizes market behavior in which firms compete between incompatible systems. Given the market with network effects, firms have natural tendencies to promote their product to de facto standardization, i.e., the existence of a single standard in the industry (Michael Katz and Carl Shapiro, 1985, 1994; Joseph Farrell and Garth Saloner, 1986a). However, a de facto standard is not overwhelmingly desirable for society. Consumers with homogenous preferences may be biased toward a new standard (Paul De Bijl and Sanjeev Goyal, 1995). A de facto standard constrains heterogeneous consumers' choices since variety in systems gives rise to higher total value for consumers with different preferences. This may reduce the degree of tipping in network markets. The coexistence of the IBM compatible and Apple in the personal computer market is a good example. There is a trade-off between variety and compatibility as explored by Joseph Farrell and Garth Saloner (1986b) and Oz Shy (1996). In addition, variety in systems is also important to avoid being stranded in an inferior standard when future technologies are uncertain (Pil Jay Choi, 1994a).

If systems are composed of components such as in a hardware-software paradigm, the market is characterized either by composite systems competition (competition between final composite product), components competition (competition at intermediate stage), or combinations of the two systems. For example, Joseph Farrell et al. (1995) examine implications for efficiency and for private incentives of two different forms of competition. Nicholas Economides and Steven Salop (1992) analyze market structures for complementary networks. They show how different market structures affect prices under independent ownership and joint ownership. They also examine the incentive for vertical integration and formation of technical standards coalitions in the industry (Nicholas Economides, 1994; Nicholas Economides and Fredrick Flyer, 1995).

Intrasystem competition describes a situation whereby competition occurs within a system. Competition between generations of a system is an intrasystem competition. It is of special interest when the network product is a durable or quasi-durable good. A firm producing such a good has to induce consumers to make repeated purchases as its technology improves. Marcel Thum (1994) first advocates that some inefficiencies due to network externalities can be internalized by market forces through competition among various contracts, i.e., the simple market contract, the upgrade contract, and the service contract, for instance, in the software market He found that simple market contracts and update contracts may lead to excessive variety and to less compatibility in oligopolistic settings. In contrast, excessive technical change emerges in the service contract. He finally shows that competition of the contracts may reduce some potential inefficiencies.

Paul De Bijl and Sanjeev Goyal (1995) point out that firms face two considerations in introducing a new-generation product. One is to upgrade the existing products, the other is to replace it. An upgrade (by keeping compatibility) enables the firm to retain network effects, but this could constrain the improvement of the product. A replacement relaxes technical constraints on product design. However, it causes a loss in benefits because of the installed base. Paul De Bijl and Sanjeev Goyal (1995) further show there may exist excess momentum by considering interactions of the two forces (network effects and constraints on design).

C. Compatibility as a Strategic Role

Consumers' benefits from compatibility arise from the expansion of network size or from the enhancement of variety and interchangeability by choosing their own idiosyncratic components (Carmen Matutes and Pierre Regibeau, 1988) in complementary networks. Although producers also gain from compatibility because of economies of scale, learning effects, and technological spillovers on the cost side (Michael Katz and Carl Shapiro, 1994), incentives for firms to move to compatibility may diverge due to asymmetry in market shares, reputation, or fear of intensification of competition. These situations induce firms to use compatibility decisions as strategies in the market.

Michael Katz and Carl Shapiro (1985) show this divergence in which firms' joint incentives for compatibility are lower than social incentives and producers with larger existing installed bases or good reputations tend to oppose compatibility. This is because firms with large or small existing networks receive asymmetric benefits from expansion of networks by compatibility. Reputations are obviously less important when all products are compatible.

The degree of competition resulting in standardization should also affect firms' decisions to achieve compatibility. It is plausible that a firm may be against compatibility if it expects standardization to intensify the competition in the market because compatibility shifts competition from characteristics of goods to costs of production, or from overall packages to specific cost of individual components (Carmen Matutes and Pierre Regibeau, 1988; Nicholas Economides, 1989). In a three-stage game with two competing technologies, Michael Katz and Carl Shapiro (1986b) show compatibility may serve to reduce rather than intensify competition in the earlier stage of a product life cycle because compatibility eliminates the threat of tipping in network markets. However, moving to compatibility leads to excessive standardization, which is not socially optimum.

Nicholas Economides (1989) compares the influence of exogenous compatibility on prices and profits of competing manufacturers' components for the case of complementary networks. He finds that prices and profits will be higher under compatibility than under incompatibility. However, total social surplus could be maximized under both compatibility and incompatibility with different parameters, but both regimes result in an over abundance of varieties compared to the social optimum.

D. Market and Nonmarket Responses

Most of the existing literature points out that many markets do exhibit network externalities. These externalities arise either in the expectations of consumers who are ill-informed, coordination failure, divergence of firms' incentives to provide compatible goods, or because of sunk switching cost for users and firms. It is interesting to examine how a market responds to or internalizes these externalities.

Well-defined property rights can help to internalize some network externalities. A single owner of a network can solely capture all benefits or overcome startup and critical mass problems from its investment on expanding the network size by exclusive proprietary patents. The video game industry is a good example (Michael Katz and Carl Shapiro, 1994). This closed strategy may frighten away consumers who are in fear of being exploited in the future. A monopoly, however, internalizes some of the externalities, but deadweight loss accrues.

In markets with many producers, implicit coordination can result if producers make their choices sequentially. But some firms that are confident they will win the game are not likely to follow a sequential decision process. Firms' contracts with consumers for provision of consumer software may serve as a remedy for externalities but only to a certain degree (Marcel Thum, 1994). The use of conversion technologies such as converters and adapters can connect incompatible technologies and hence might be used as a mechanism for reducing externalities. The use of conversion technologies, however, makes things worse because converters reduce private costs without a form of standardization, and they decrease social welfare by making standardization less possible (Joseph Farrell and Garth Saloner, 1992).

Firms can internalize network externalities by contracting or by integrating vertically. A firm can contract with a large group of users (e.g., government) to overcome the startup and the critical mass problem. In a hardware-software market, a firm can subsidize its suppliers to produce complementary goods. A joint venture or industry coalitions can crush similar problems. Sony's acquisition of Columbia Pictures and Matsushita's acquisition of MCA (Michael Katz and Carl Shapiro, 1994) are cases that reflect contracts entered into modify these types of problems.

As shown in the literature, market mechanisms often cannot internalize externalities and inefficiencies arise as a consequence. However, relatively little attention has been paid to non-market responses to these problems. The exception is the study by Joseph Farrell and Garth Saloner (1988). This study is a pioneering work on how a market obtains explicit coordination through the use of standard committees. Many national as well as international private-sponsored institutions serve in the role of creating better market outcomes in practice. Examples of such standardization committees include the American National Standards Institute (ANSI) and the International Organization for Standardization (ISO).

In finite and infinite period games, Joseph Farrell and Garth Saloner (1988) consider relative speed, efficacy, and payoffs of two firms that seek a choice between two incompatible technologies. Their model shows that a committee system outperforms the bandwagon system, in which an agent makes a unilateral commitment to one standard and others follow the lead, but the differences in payoffs vanish as the number of periods goes to infinity. A committee solution is likely to take a long time while a bandwagon system may have an early outcome since no firm is likely to yield until the deadline set by the committee. However, the bandwagon system may not necessarily produce the standardization outcome.

Government may subsidize network industries to internalize the network effects as indicated by Philip Dybvig and Chester Spatt (1983) and Michael Katz and Carl Shapiro (1994). Some pioneering research work has been subsidized. For example, the development of the Internet and the TCP/IP protocol were initially government projects subsidized by the U.S. Defense Department in the late 1960s and the National Scientific Foundation (NSF) in the mid-1980s (Jeffrey MacKie-Mason and Hal Varian, 1994a).

A third method to remedy this unfavorable situation is through direct regulation by government. Little research on government regulation exists, however. In practice, government is involved in network markets. For example, the Federal Communications Commission (FCC) regulated color televisions in 1950s. Regulation of High Definition Television (HDTV), recent consideration on alternative Digital Television (DTV) standards, and the European Community's regulation of HDTV (Michael Katz and Carl Shapiro, 1994; FCC, 1996) are also prime examples.

Michael Katz and Carl Shapiro (1994) point out that several issues should be noted before concluding that government regulation is a necessity: (1) market inefficiency is unclear, (2) government incentives to improve outcomes may place higher weight on the current generation and hence impose inefficiencies on emerging technologies, and (3) government may have imperfect information upon which to base regulation. By borrowing ideas from political economy models (Sam Peltzman, 1976; Gary Becker, 1983; Ngo Van Long and Neil Vousden, 1991; Gene Grossman and Elhanan Helpman, 1995a, 1995b), incentives for and the outcome of government intervention in network markets are taken up in the third essay.

Ronald Coase (1972) claimed that the price of a durable good produced by a monopolist who cannot commit to restricting output will converge to marginal cost quickly. A declining price path should be observed in such a monopoly market unless a commitment mechanism exists or unless lease contracts are allowed (Jeremy Bulow 1982, 1986; Lawrence Ausubel and Raymond Deneckere, 1989). Michael Waldman (1993) and Pil Jay Choi (1994b) introduce network externalities to study their effects on the monopoly market. They both focus on the decision of compatibility between successive generations of products rather than pricing. It is of particular interest to investigate if upgrades and network effects help to modify the downward pressure on the price path faced by the monopolist. Upgrade decisions specifically related to the time consistency paradox raised by Ronald Coase (1972) and subsequently by Jeremy Bulow (1982, 1986) are taken up in Essay II, titled Upgrades, Commitment, Monopoly Profits, and Network Effects. A game-theoretic setting is used to determine if upgrades and the presence of network effects offer the monopolist commitment strategies that can be used to moderate the downward pressure on prices, i.e., modify the adverse conclusions or the Coase conjecture.

The question whether or not government should intervene in network markets is still debatable, especially when government involvement increases (e.g., the standard regulation for HDTV and DTV). The incentives of government to improve market outcomes are challenged by many political economy researchers. Regulatory outcomes are interesting and different when the government is assumed to be a political-welfare maximizer rather than a social welfare maximizer (Sam Pelzman, 1976; Gary Becker, 1983; Ngo Van Long and Neil Vousden, 1991; Gene Grossman and Elhanan Helpman, 1995a, 1995b). This is a controversial subject but has received little attention in the literature even though there is increasing involvement on the part of government in such markets. This issue is taken up in Essay III, titled Political Economy of Network Standardization, which considers the case for government intervention in network markets. The model developed in this essay attempts to predict the outcome of standard setting by government intervention under the assumption that the government agent maximizes political welfare rather than social welfare.

Essay IV, titled Rational Incompatibility with International Product Standards, extends Michael Katz and Carl Shapiro's model (1985) to international markets. This essay considers the private incentives of an import-competing firm and the social incentives of a technology-importing country to achieve compatibility for a product characterized by network effects. I find that the social incentive of an importing country to deviate from the international standard is even greater than the private incentives, providing, in turn, incentives to adopt and enforce technical barriers to trade. These results confirm the problems of path dependence and may offer some insights for the international standard-setting process.