The interest in network economics, where network effects, compatibility, and standards play a predominant role, has risen enormously over the last decades. It gave rise to a large body of theoretical literature, which showed that the existence of strong network effects altered market outcomes in many important ways, often leading to market failure. Examples of such market failure include market breakdown due to start-up and hold-up problems, as well as inability of industries to switch to a superior standard, known as “excess inertia”.
In fact, these topics have also attracted a lot of attention outside economics, including business scientists and practitioners, as well as policy makers. One reason for this interest is that the network aspects are more apparent in the fast growing and widely reported industries – the information industries – than in more traditional ones. The famous antitrust cases of AT&T, IBM, and Microsoft contributed to the popular interest in the information industries, as well as challenged theoretical and empirical economists.
It has been claimed that, since we are living in a “New Economy”, a “New Economics” is needed to understand its functioning and to guide business strategy and public policy. Although, this claim seems exaggerated (Shapiro and Varian, 1999, p. ix-x), there is definitely a need for more research in network economics. In particular, this applies to empirical research, since quantification of network effects and compatibility in real markets is still a largely uncovered issue.
In general, demand for a good exhibits network effects, if the utility of each consumer increases with the number of other consumers purchasing the same good. The notion of network effects is very closely related to – and in fact often interchangeably used with – the notion of network externalities, positive feedback, and demand-side economies of scale.
Economic theory predicts that network effects can lead to multiple equilibria and particular market shares’ dynamics. Multiple equilibria may result even if the network good is offered by a single provider. The seminal contribution in the field is the work by Rohlfs (1974), who studies demand for telephone service. He shows that a monopolist offering the service faces multiple outcomes in terms of equilibrium user sets (network size) at a given price. He then points out the start-up problem for a new communications service stressing that a critical mass of users is needed in order to secure the large network size. In dynamic models, overcoming of the critical mass is reflected by an S-shaped diffusion of network good adoptions (Cabral, 1990).
Katz and Shapiro (1994) provide an overview of market responses to the start-up problem in communication networks discussing various pricing and expectations-management strategies. They extend the discussion to system networks, which consists of two components consumed together: hardware and software. In such systems, indirect network effects can arise, as opposed to direct effects in communication networks. Indirect effects arise, if utility of the hardware users increases with the number of other hardware users, because it positively influences quality and variety of the available software. Providing such systems is potentially vulnerable to hold-up problem, since the buyers of durable hardware could expect to be exploited by the future software price increases (Farrell and Gallini, 1988). Under-utilization of the communication networks and the system networks due to the start-up and hold-up problems is one example of market failure in provision of network goods. In extreme cases (e.g. pure network goods), such under-utilization can lead to complete market breakdown.
Another example of market failure studied in the literature on network economics is “excess inertia” or lock-in by historical events (Farrell and Saloner, 1985; Arthur, 1989). A particularly entertaining reading in this strand of literature is David (1985) in which the author illustrates how QWERTY keyboard standard won the competition with more efficient Dvorak keyboard standard. Industries can be locked-in in an inferior standard, because network [Seite 3↓]markets are prone to “tipping”, which is the tendency of one standard to drive the competitors out of the market once it gained a lead. Tipping occurs in static (one-shot) models in the form of multiple equilibria in which a single standard dominates (Katz and Shapiro, 1985). In dynamic models, tipping is reflected in equilibria where the adoptions of the loosing standard die out once a rival standard is introduced or accepted in the market (Farrell and Saloner, 1986a; Katz and Shapiro 1992, 1994). Tipping does not necessarily, however, lead to excess inertia in introducing new superior standards, as stressed in the earlier works. Indeed, properly defined property rights could alleviate or even reverse the excess inertia problem (Katz and Shapiro, 1986a, 1992; Liebowitz and Margolis 1990).
The above mentioned properties of markets featured by network effects give rise to particularly rich array of strategies and tactics on the side of firms, especially when compatibility is a decision variable (Katz and Shapiro, 1994; Farrell and Klemperer, 2001). Moreover, evaluation of these strategies and tactics by anti-trust authorities is a much more delicate issue in network industries than in more traditional industries (Economides and White, 1994; Röller and Wey, 2003). For instance, in contrast to traditional industries, marginal cost pricing may result in market failure due to the start-up problem in network industries. Hence, introductory pricing below marginal costs may be necessary as a means to achieve large network size. Yet, it is difficult to distinguish introductory pricing from anti-competitive predatory pricing in practice. At the same time, market dominance does not necessarily imply that the market leader earns super-normal profits. The quasi-rents may reflect the costs incurred earlier in order to attract the critical mass of consumers.
An important factor conditioning the impact of network effects is compatibility. If the goods are perfectly compatible, consumers of all the goods make up a common network and contribute to the network effects at the industry level. In the polar case of perfect incompatibility, consumers of each good form separate networks and the network effects operate at the firm level.
The importance of compatibility between competing network goods for performance of network industries is twofold. First, it directly influences the gross consumption benefits of consumers. They directly benefit from compatibility since it expands the size of each network to the total membership of both. The drawback of compatibility, however, is a loss of variety (Farrell and Saloner, 1986b). In general, we might expect that with strong network effects any loss of variety is a minor price to pay to achieve compatibility, as in the case of VHS vs. Beta in videocassette recorders (Cusumano et al., 1992). Second, compatibility alters the nature of [Seite 4↓]competition between the network good providers thereby indirectly influencing the benefits of consumers. As stated by Besen and Farrell (1994), the providers compete “for the market” under incompatibility, whereas under compatibility they compete in a more traditional manner “within the market”. In other words, winning or monopolizing the market will more often occur under incompatibility. The reason behind it is that positive feedback at the firm level (incompatibility) makes network markets more prone to tipping than positive feedback at the industry level (compatibility). A natural question to ask then is how compatibility affects the overall degree of competition between the providers. There is no simple answer to it. Compatibility tends to relax competition early in the product-life cycle, because the threat of tipping is reduced. However, it tends to intensify competition later on, because it makes monopolization of the market less likely (Katz and Shapiro, 1986b).
Empirical research on network effects confirms their importance in real markets. The following brief overview distinguishes between descriptive and structural econometric analyses (Reiss and Wolak, 2002). The difference is that the former proceeds without reference to an economic model, whereas the essential components of the latter are the theoretical and statistical assumptions that allow a researcher to recover economic primitives from data. As a consequence, estimates of economic magnitudes and the extent of causation can be recovered solely from structural analyses.
The early wave of descriptive econometric studies confirmed that compatibility matters in network markets, including mainframe computer systems (Greenstein, 1993) and computer software (Gandal, 1994, 1995; Brynjolfsson and Kemerer, 1996). Further, empirical evidence of network effects was found in the adoption of automated teller machines (Saloner and Shepard, 1995), microcomputer systems (Gandal, Greenstein, and Salant, 1999; Goolsbee and Klenow, 2002), mobile telephone service (Okada and Hatta, 1999; Kim and Kwon, 2003), and DVD players (Dranove and Gandal, 2003).
The small but growing structural econometric literature focuses on identification of indirect network effects, which arise in system networks. Typically, the identification is achieved by estimating demand for both hardware and software simultaneously and testing for interdependence between them. Substantial extent of indirect network effects was found in markets for compact disc players (Gandal, Kende, and Rob, 2000) and video cassette recorders (Ohashi, 2003). Moderate extent of the effects was reported in Yellow Pages market (Rysman, 2002).
Structural econometric modeling of demand exhibiting direct network effect is still a largely uncovered issue. Economides and Himmelberg (1995) conducted the pioneering study in this field. They estimated demand in the market for facsimiles assuming perfect competition on the supply side. The parameters of utility function that they recovered from the estimation results suggest extremely strong network effects in this market. Our work builds on the Economides and Himmelberg (1995) carefully examining the identification issues and introducing imperfect competition into the model.
The thesis proposes a structural econometric model of demand exhibiting direct network effects. The structural approach we follow allows us to identify the extent of network effects and compatibility between competing networks. To the best of our knowledge, this is the first model that facilitates the identification in the case of direct network effects. At the same time, it is the first attempt to assess the degree of compatibility in an econometric framework. The model is then applied to investigate demand for mobile telecommunications service and the link between adoptions of ISO 9000 standard and international trade. The estimation results allow us to formulate some interesting policy conclusions.
The first paper of this thesis, “Identification of Network Externalities in Markets for Non-Durables”, introduces the economic structure of the model. Given the economic primitives, we derive the equilibrium diffusion path of the network goods’ adoptions. There are two major assumptions that drive the identification of network effects and compatibility in the model. First, network effects are captured by lagged network size in the consumer willingness-to-pay function. Second, providers of the competing network goods set equal hedonic prices.
The first assumption assures uniqueness and continuity of the equilibrium diffusion path. Using lagged instead of current network size in the willingness-to-pay function, we implicitly rule out coordination between consumers. Without coordination, the adoptions follow minimum equilibrium diffusion path, which still will be discontinuous if the lag is infinitely small. In practice, the frequency of data used to estimate the model will determine the lag thereby making the diffusion path continuous. We also discuss the implications of the lag for the extent of identified network effects and rationality of consumers in the model.
The second assumption means that prices “corrected” for the network size are equal across the competing network goods. Hence, without other quality characteristics, which affect value of the goods, nominal (“uncorrected”) prices that are unequal across the goods imply the existence of network effects and incompatibility. At the same time, network effects, as captured by the first assumption, trigger the impact of lagged network size on current network size. This inertia in network size dynamics together with the nominal price differential across goods identifies the extent of network effects and compatibility in the econometric model.
We apply this model to investigate demand for mobile telecommunications service in the second paper of this thesis, “Estimating Network Effects and Compatibility in Mobile Telecommunications”. The firm-level data we use stems from the Polish mobile telephone industry and covers the period 1996-2001. Our empirical analysis points to strong network effects and, despite full interconnection of the mobile telephone networks, low compatibility. These findings suggest intra-network call discounts and transmission of information on quality as main sources of network effects in mobile telecommunications. Finally, we show that ignoring network effects causes overestimation of price elasticity of demand. One policy conclusion from this is that subsequent measure of market power will be overestimated as well.
The last paper, “Diffusion of ISO 9000 Standards and International Trade”, investigates the impact of ISO 9000 adoptions on bilateral trade flows. The ISO 9000 family of standards is designed to provide confidence to buyers that products will meet their expectations thereby enhancing trade and global welfare. However, its critics claim that it is merely a barrier to market entry and a tariff on international trade. Using a panel data set on 101 countries over 1995-2001 we estimate a gravity equation for bilateral exports in which domestic and foreign ISO 9000 adoptions affect bilateral trade barriers. One econometric caveat there is that ISO 9000 adoptions could be endogenous. The first paper of this dissertation offers a way to endogenize the adoption decisions. The basic idea being that ISO 9000 is a common language lowering informational asymmetries between firms. Consequently, we expect value of these standards to increase with the number of adopters giving rise to network effects. Although, formal tests do not reject exogeneity of regressors in our trade equation, we still estimate a diffusion equation of ISO 9000 in order to obtain additional evidence of the role of these standards in international trade.
In general, our empirical results are consistent with the common language hypothesis. We find that the adoptions of ISO 9000 benefit bilateral trade, as well as spur further domestic and foreign adoptions. At the same time, we observe a substitution effect, as certified firms tend to trade more with each other than with uncertified firms. In fact, this creates an effective barrier to trade for less developed countries. The substitution effect works against them, because adoptions of the standards are concentrated in more developed countries. One policy conclusion we draw from this is that the benefits of ISO 9000 will remain in the more developed countries domain unless we find and remove the barriers to its global diffusion.
|© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.|
|XDiML DTD Version 4.0||Zertifizierter Dokumentenserver|
der Humboldt-Universität zu Berlin
|HTML-Version erstellt am: