105 justified. Rather, market institutions are creative in finding ways by which consumer preferences are revealed, such that network externalities can be internalized within regular market operations. Policy makers would find it difficult to improve on market outcomes. With respect to the allocative efficiency of the market outcome to be expected in an unregulated contestable market featuring network externalities, it can be said that, since the cost conditions of the market favor only one active firm, there is no threat of network islands developing. Furthermore, should the operator try to enforce a price above the competitive level, this would invite market entry. Only the competitive price protects the operator from replacement by a potential competitor. This competitive price level leads to a network size that internalizes the network externalities to a large degree (comparable to equilibrium Sp in Figure 6.1). 6.3.3 Case 3: Network externalities in a competitive market In a competitive market featuring substantial network externalities, firms will compete not only in price and product characteristics, but also in the dimension of network size. The stronger the network externalities, the more important will be the “network effect” a product offers compared to its “technology effect” (see section 6.1 above). The important difference to a contestable monopoly with network externalities is, that in a competitive market users have a choice between several operating networks. The trade-off between standardization and variety, mentioned in section 6.2, takes on its full meaning only in this context. An increased variety of technologies available for consumption makes it more difficult for consumers to resolve the trade-off between choosing the technology with the most preferred product characteristics and benefiting from being a part of a network with a large number of users. Generally, consumers can be expected to have different preferences with respect to their preferred technological and qualitative product characteristics and also different valuations of the network effect. It is therefore likely that consumers are better off when a variety of network islands caters to particular consumer tastes compared to one large network offering a compromise between the demanded product characteristics (Shy, 2001: 27f.). However, the more important the network effect, the more willingly will consumers give up specialized technologies and concentrate on the network offering the largest user base. The aim of this section is to understand how the competitive market process will solve the trade-off between taking advantage of network externalities (implicating a small number of active firms) and offering product variety and competition in the market (implicating a larger number of active firms). In contrast to the case of the trade-off between economies of scale and product variety, there is a solution to the trade-off between network externalities and product variety. Consumers can profit both from product variety and the positive network externalities when rival firms are willing to cooperate by making their networks 106 compatible. Compatibility between otherwise self-contained network islands allows users to communicate and share network benefits across networks, while competition is preserved in product dimensions not affected by the cooperation.95 With full compatibility the consumer will base his product choice on technological characteristics and price. Network reach will be the same for all products. The main question for the present analysis is what incentives firms in a competitive market have to make their networks compatible. Firm incentives for compatibility When network externalities are strong and the consumer valuation of compatibility accordingly high, the question is whether firms are likely to voluntarily engage in cooperation while salvaging competition in product characteristics. Or is it more likely that larger firms will decline cooperation in the hopes of gaining a competitive advantage from offering the largest network? Can a larger network charge higher prices and make above competitive profits in competitive markets with strong network externalities? To answer this question it is useful to consider the costs and benefits of cooperation from the perspective of the firm. Most substantial on the cost side is that, by making its network compatible, a firm gives up network size as a differentiating product characteristic in its competition with rival firms. This aspect is especially costly to firms with initially large networks. Furthermore, cooperating firms have to invest more in product differentiation with respect to remaining differentiating characteristics. Lastly, there are transaction costs involved in coordinating the cooperation of firms with conflicting interests. The most substantial of the positive aspects of compatibility is the possibility to benefit from a significant demand shift generated by the increase in overall network size. When total demand for network services rises due to several networks becoming compatible, then even initially large networks can experience a demand increase for their network services. Economides (1996), for instance, shows that even a monopolist can profit from inviting entrants, even subsidizing their market entry, and making his services compatible to theirs. This counterintuitive result can nonetheless be explained easily: when consumer demand is a function of the expected network size, and when network externalities are sufficiently strong, then the demand effect from a larger network pushes the demand curve outwards to such an extent that the higher market-wide sales allow the monopolist to sell so much more of his product compared to the monopoly output that, even at a lower price, he can 95 Put differently, supply-side scale economies determine the cost-efficient physical network size, while user externalities determine the efficient number of compatible network users. Since physical network size cannot be shared between independent network owners, supplyside economies of scale lead to higher firm concentration in a market. User externalities do not have this inevitable effect, since the “user size” of a network can be shared through cooperation. 106 compatible. Compatibility between otherwise self-contained network islands allows users to communicate and share network benefits across networks, while competition is preserved in product dimensions not affected by the cooperation.95 With full compatibility the consumer will base his product choice on technological characteristics and price. Network reach will be the same for all products. The main question for the present analysis is what incentives firms in a competitive market have to make their networks compatible. Firm incentives for compatibility When network externalities are strong and the consumer valuation of compatibility accordingly high, the question is whether firms are likely to voluntarily engage in cooperation while salvaging competition in product characteristics. Or is it more likely that larger firms will decline cooperation in the hopes of gaining a competitive advantage from offering the largest network? Can a larger network charge higher prices and make above competitive profits in competitive markets with strong network externalities? To answer this question it is useful to consider the costs and benefits of cooperation from the perspective of the firm. Most substantial on the cost side is that, by making its network compatible, a firm gives up network size as a differentiating product characteristic in its competition with rival firms. This aspect is especially costly to firms with initially large networks. Furthermore, cooperating firms have to invest more in product differentiation with respect to remaining differentiating characteristics. Lastly, there are transaction costs involved in coordinating the cooperation of firms with conflicting interests. The most substantial of the positive aspects of compatibility is the possibility to benefit from a significant demand shift generated by the increase in overall network size. When total demand for network services rises due to several networks becoming compatible, then even initially large networks can experience a demand increase for their network services. Economides (1996), for instance, shows that even a monopolist can profit fro inviting entrants, even subsidizing their market entry, and making his services compatible to theirs. This counterintuitive result can nonetheless be explained easily: hen consumer demand is a function of the expected network size, and when network externalities are sufficiently strong, then the demand effect from a larger network pushes the de and curve outwards to such an extent that the higher market-wide sales allow the monopolist to sell so much more of his product compared to the monopoly output that, even at a lower price, he can 95 Put differ ntly, supply-side scale economies determi e the c st-efficient physical network size, w ile user exte nal ties determine th efficient number of compatible network users. Sinc physical network size cannot be shared between independent network owners, supplyside economies of scale lead to higher firm concentration in a market. U er externalities do not have this inevitable effect, since the “user size” of a network can be shared through cooperation. 107 increase his profit.96 The former monopolist can realize this output level and this price only in a competitive setting. As a monopolist he cannot credibly commit to producing a competitive output/price combination because, given a level of demand, it is always in the monopolist’s interest to restrict output and raise prices. Since consumers expect the monopolist to offer network services only at higher prices, they expect a smaller network in a monopoly and their demand at any given price will be correspondingly lower in a monopoly than in a competitive market. A further benefit to cooperation is that it eliminates competition to establish the largest network and therefore frees resources, which can be put to other uses. These can, for instance, be invested into competing in other product characteristics. The more important technological differentiation is to consumers, the more important it will be to invest into product innovations. It can be shown that, ceteris paribus, firm incentives for compatibility are higher in markets characterized by product differentiation (Farrell and Saloner, 1992; Doganoglu and Wright, 2006). Since product innovations are likely to be of greater significance in dynamic markets, such markets may also be more likely to show compatibility among firms. A model of network externalities in Cournot competition There is no reference model for the case of network externalities in competitive markets. Models dealing with several competitors in markets featuring strong network externalities generally do not allow for free market entry. Rather they model compatibility choice in Cournot competition between a fixed number of firms. The present section discusses why the results of these models cannot be transferred to a free-entry market environment. The seminal paper on network externalities in oligopolistic competition is Katz and Shapiro (1985). The authors analyze equilibrium outcomes and firm incentives for compatibility. The model defines a consumer’s utility as: Ui,j(S,T) = ri + v(Sje) for consumers i = 1,2,...m and technologies j = 1,2,...n. Parameter ri stands for the basic willingness to pay for network services (this is the willingness to pay which results from the “technology effect”). As technologies are assumed to be homogeneous in all characteristics but network size, r does not differ across technologies j, but only across consumers i, assumed to have different basic valuations of network services. The parameter r is assumed to be uniformly distributed between minus infinity and A with density one, and A is assumed to be a positive value.97 The consumer’s valuation of the network externality is expressed in the externality function v(Sje), where Sje is the sum of the network users expected to be compa- 96 The elasticity of demand obviously plays an important role in this context. The more elastic the demand, the more pronounced will be the demand expansion induced by more rivalry in the market. 97 These assumptions on the distribution of r ensure a linear demand curve for technology j. 108 tible with technology j. All consumers have the same valuation of the network effect. The model assumes v‘(Sje) > 0, v‘‘(Sje) < 0 and lim v‘(Sje) = 0 as v‘(je)??. Users counted among Sje are either subscribers of technology j or subscribers of networks compatible with technology j. When xke stands for the expected network size of technology k, Sje can be expressed as: where k = 1,2,...m are the networks compatible with j. The static, one period model is solved in three steps. In a first step consumers build expectations about the future network size associated with each technology. In a second step firms choose their output given the consumer expectations of network size and assuming the output of rival firms as given (Cournot competition).98 This output game determines the prices that firms charge in equilibrium. In a third step, consumers make their purchase decision by maximizing their utility. When the price of a technology is given by pj, a consumer chooses the technology for which utility is maximized (and is equal to or greater than the utility realized with the next best alternative resource utilization). Formally, the consumer’s decision function is: Assuming that Umin is equal to zero, the consumer maximizes If this expression is negative for all available technologies j the consumer will join none of the available networks. Since the consumer’s basic willingness to pay ri is the same for every available technology j, the consumer will choose that technology for which the difference [v(Sje) - pj] is maximized, respectively that technology for which the price adjusted for the network effect offered by a particular technology, [pj - v(Sje)], is minimized. Katz and Shapiro call this price the “hedonic price” of technology j (Katz and Shapiro, 1985: 427). Since all consumers are assumed to have the same evaluation of the network externality, all consumers will choose the technology that is available at the lowest hedonic price. Therefore, the equilibrium will only feature more than one firm with positive output, if these firms offer their technology at the same hedonic price. 98 The model assumes constant marginal costs. ( )[ ] min, ,max UpTSU jjij ?? ( ) ( ) jejijji pSvrpTSU ?+=?,, ? = = m k e k e j xS 1 109 The equilibrium concept employed in the model is that of fulfilled expectations Cournot equilibrium (FECE). This concept assumes that all firms maximize their profits by choosing their output. Consumer expectations and the output level of rival firms are given. Production costs are assumed to be the same for all firms; all costs are non-variable. In equilibrium expected sales must be equal to actual sales. Assuming that fixed costs are zero and redefining ri as the willingness to pay which exceeds marginal costs, allows the authors to set up a profit function, which, differentiated with respect to output, yields output equations for all firms. Solved simultaneously, these equations yield the following equilibrium output: Since the number of active firms n is exogenous to the model, the model leads to the conclusion that those networks which consumers expect to be dominant (high value of v(Sje)), for instance because of their initial network size or because of their reputation, will in fact be dominant in equilibrium (high xj*). Therefore, within their framework, the authors conclude that network externalities lend market power to initially large firms. Discussion of the model The equilibrium outcome that those networks that consumers initially expect to become dominant will in the end dominate the market is a result of the specifications of the model. Particularly, the model does not include a positive value to product variety. For consumers to value variety, the parameter r, reflecting the willingness to pay for the technological characteristics, would have to vary in technologies j. If this were the case, then not all consumers would prefer the same network, even if their valuation of the network externality were the same. Consumer expectations of network size would then not necessarily dominate the consumer’s choice of network. Furthermore, the assumption that all consumers have the same valuation of the network effect is unrealistic, and at the same time very restrictive with respect to the results of the model. Lifting these assumptions would allow different combinations of price, technological characteristics, and network size to exist simultaneously in the market. The result that networks that have an initially large user-base will come to dominate a market would no longer hold. More realistic market conditions (heterogeneous consumers, product differentiation, and free market entry) are likely to lead to more firms in equilibrium. Katz and Shapiro analyze how the result would change if the assumption of a fixed number of firms is lifted and instead the number of firms is assumed to increases indefinitely. The equilibrium output would then converge on: ( ) ( )1* +÷?? ? ?? ? ??+= ? ? nSSvnAx jk e k e jj 110 In words, as the number of firms grows, firm j’s equilibrium output will be the larger, the larger the expected network size of firm j is compared to the network sizes of its competitors. When firms are expected to have the same network size in equilibrium, either because of full compatibility in the market or because firms are grouped in network islands of equal size, then the equilibrium output of each single firm will tend towards zero (because the second term of the above equation, will tend towards v(Ske), and v(Ske) is equal to v(Sje), when networks are symmetric). Using this result, Katz and Shapiro (1985: 429) show that when there is full compatibility between all active firms, then there is a unique symmetric equilibrium, which converges on the perfectly competitive equilibrium, when the number of firms becomes increasingly large. This competitive case analyzed by Katz and Shapiro is trivial. Since the model does not allow competition by product differentiation the firms are homogeneous whenever there is full compatibility. Compatibility in this setting takes away the last differentiating aspect, namely network size. Naturally a market with an infinite number of homogeneous firms will converge on the perfectly competitive equilibrium. A realistic model of competition, which would allow different consumer valuations of technological characteristics and network size, would not predict perfect competition in a competitive market featuring network externalities. This is, however, as was already stated above, not the criterion against which market performance should be measured. A more promising approach for understanding firm incentives for compatibility in competitive markets is to analyze the difference in realized profit from the noncompatibility case to the compatibility case. Katz and Shapiro (1985) show that within their framework not all suppliers benefit from increased compatibility. Even when overall output rises, initially large networks may have a lower output in a compatible equilibrium. When the possibility that firms make side-payments among one another is considered, then it suffices that average profits in the market increase as a result of cooperation for firms to voluntarily agree on compatibility. Those firms that gain disproportionately from compatibility can compensate the losses of those firms that would otherwise prefer not to cooperate.99 Because the condition that average profits rise for compatibility to be a viable outcome is less restrictive than the condition that profits increase individually, side-payments increase the 99 There are strong parallels here to the Coase Theorem, which postulates that as long as property rights are well-defined, private negotiations can lead to efficient market outcomes. See also Katz and Shapiro (1986: 825). ( ) ( ) ( )?????? +÷?= ?????? jk eknejjn nSvSvx 1limlim * 110 In words, as the number of firms grows, firm j’s equilibrium output will be the larger, the larger the exp cted network size of firm j is compared to the network sizes of its competitors. Wh n firms are expected to have the same ne work size in equilibrium, eith r because o full compatibility in the market or because firms are g ouped n network islands of equal size, then the equilibrium output of each singl firm will tend t wards zero (because the s cond term of the above equ tion, will tend towards v(Ske), and v(Ske) is equal to v(Sje), when networks ar symmetric). Using this result, Katz and hap ro (1985: 429) sho that when there is full compatib lity between all ctive firms, then there is a unique symmetric equil bri m, which converges on the perfectly competitive equilibrium, when the number of firms bemes increasingly large. This competitive case analyzed by Katz and Shapiro is trivial. Since the model does not allow competitio b product differentiation the firms are homogeneous whenever there is full compatibility. Compatibili y in this setting t k s away the last differ ntia ing a pect, namely network size. Naturally a market with an infinite numb r of homogeneous firms will converg on the perfectly competitive equil brium. A realistic mod l of co petition, which would allow different consumer valua tions of technological charact r stics and network size, would ot predict perfect c mpetition in a competitive market featuring network externalities. This is, however, as was already stated above, not he criterion against which market perfor mance hould be measured. A more promising approach for understanding firm incentives for compatibility in competitive markets is to nalyze the difference in realized profit from the nona bility cas to the compatibility case. Katz and Shapiro (1985) show that within their fram w rk not all suppliers benefit from increased compatibility. Even hen ov rall output rises, initially la ge networks may h ve a lower output in a compatible equilibrium. When the possibility that fir s make side-paymen s among one anoth r is cons dered, then it suffices that average profits in the market increase as a result of cooperation for firms to voluntarily agree on compatibili y. Tho firms that gain dis roportionately from compatibility can c mpensate the losses of those firms that would otherwise prefer not to cooper te.99 Because the c nditi n at average profits rise for compatibility to be a viable outcome i less restric ive n the condition that profits increase individually, side-payments increase th 99 There are strong parallels here to the Coase Theorem, which postulates that as long as property rights are well-defined, private negotiations can lead to efficient market outcomes. See also Katz and Shapiro (1986: 825). ( ) ( ) ( )?????? +÷?= ?????? jk eknejjn nSvSvx 1limlim * 111 likelihood that firms reach a mutually beneficial compatibility agreement, even when they are of substantially different initial size.100 Public policy for competitive markets featuring network externalities The above discussion shows that cooperation is likely in competitive markets with substantial network externalities when (1) consumer tastes vary, (2) when firms have the possibility to differentiate their services (not only in network size, but also along technological and qualitative characteristics) and (3) when compensation for compatibility is possible. Side-payments should not be interpreted as reflecting market power. Rather, they redistribute the gains from network compatibility in a way that all firms are better off. In conclusion, there are many reasons for firms in competitive markets to voluntarily agree on compatibility when strong network externalities are present. It is not possible to derive a general justification for interventions by competition policy in the presence of network externalities. If competition authorities consider it necessary to intervene into a competitive market on the grounds of market power caused by strong network externalities, they need to prove for this particular case that the above mentioned prerequisites for compatibility are not fulfilled, and that there exists a large firm which is abusing its market position, earning above competitive profits. 6.3.4 Compatibility standards Before network services can become compatible across network boundaries technological requirements will have to be fulfilled. How do firms decide on a standard to which all compatible networks have to conform? One can differentiate between three general mechanisms by which compatibility between firms can be reached. The first is standardization by multilateral agreement. For this the firms convene in standardization committees charged with reaching a consensus on a particular product standard. The second mechanism is a unilateral predetermination of a particular standard by a leading firm in the industry. The remaining firms either adopt this standard or employ adapters or gateway technologies in order to make their products compatible to the standard set by the industry leader. Lastly, an industry standard can be imposed by governmental decree. 100 Compensation here refers strictly to payments made in order to induce the other party to agree on a common standard. These payments have to be differentiated clearly from payments which are compensation for costs incurred by the other party as a consequence of the compatibility. For example, if one network offers another network free licenses for its software products in order to induce it to agree to interconnect this can be considered a side-payment. When, however, payments flow in direct relation to the costs of achieving compatibility, such as transportation compensation for the costs incurred in terminating one another’s traffic, then these payments do not reflect side-payments.