125 are interested in a good quality interconnection because all profit equally from a demand-expansion effect. The elicitor of a quality degradation would suffer the same negative demand reduction as its three rivals without a compensatory gain from a positive quality-differentiation effect. The authors then show how the incentives to interconnect change when two of the ISPs merge and the resulting market of three ISPs then includes one firm with an installed base of at least the size of the combined installed bases of the other two firms. In this scenario the largest firm is not interested in deteriorating the quality of interconnection with both of the rival networks. However, in some circumstances, it can profit from a targeted degradation strategy, in which it refuses good quality interconnection with one of the smaller rivals while it continues good quality interconnection with the other rival. This conclusion depends on the non-targeted firm not offering transit services to the targeted firm.111 The positive quality-differentiation effect will then result in the targeted firm not attracting any new customers while the dominant firm and the non-targeted firm gain more customers (even though the non-targeted rival profits more from the quality-differentiation effect). It was especially this result that competition authorities relied upon in their decision on the merger by MCI and Worldcom in 1998. 7.3.2 Critique of the model by Crémer, Rey and Tirole and alternative modeling In chapter 6 it was discussed that the results of Katz and Shapiro (1985) depend critically on the additional assumptions in the model besides the network externalities. These additional assumptions introduce elements of market power such that the model cannot be used to prove that network externalities lend market power to large network operators. The same holds true for the adaptation of the Katz and Shapiro model by Crémer, Rey and Tirole. Here also there are significant additional assumptions in the modeling set-up which lead to the result that the largest firm prefers a lower level of interconnection quality compared to its smaller rivals. Below it is discussed whether these assumptions are relevant for the market for Internet backbone services. Market entry conditions First, consider the assumption of a fixed number of firms in the market. This assumption does not correspond well to the thousands of active ISPs observable in reality. If at all, then this assumption may apply to the market for Tier-1 ISP services in which only 5 to 10 ISPs are active. But whether this market has structural barriers 111 Crémer, Rey and Tirole (ibid., 458) argue that the dominant firm can limit the capacity of the interface with the non-targeted network to such an extent that the capacity is only sufficient to provide good quality interconnection for the traffic of the non-targeted network but would result in very bad interconnection quality if the traffic should grow to encompass also the traffic of the targeted network. 125 are interested in a good quality interconnection because all profit equally from a de and-expansion effect. The elicitor of a quality degradation would suffer the same negative demand reduction as its three rivals without a compensatory gain from a positive quality-differentiation effect. The authors then show how the incentives to interconnect change when t o of the ISPs merge and the resulting market of three ISPs then includes one firm with an installed base of at least the size of the combined installed bases of the other two firms. In this scenario the largest firm is not interested in deteriorating the quality of interconnection with both of the rival networks. However, in some circumstances, it can profit from a targeted degradation strategy, in which it refuses good quality interconnection with one of the smaller rivals while it continues good quality interconnection with the other rival. This conclusion depends on the non-targeted firm not offering transit services to the targeted firm.111 The positive quality-differentiation effect will then result in the targeted fir not attracting any new customers while the dominant firm and the non-targeted firm gain more customers (even though the non-targeted rival profits more from the quality-differentiation effect). It was especially this result that competition authorities relied upon in their decision on the merger by MCI and Worldcom in 1998. 7.3.2 Critique of the model by Crémer, Rey and Tirole and alternative modeling In chapter 6 it was discussed that the results of Katz and Shapiro (1985) depend critically on the additional assumptions in the model besides the network externalities. These additional assumptions introduce elements of market power such that the model cannot be used to prove that network externalities lend market power to large network operators. The same holds true for the adaptation of the Katz and Shapiro model by Crémer, Rey and Tirole. Here also there are significant additional assumptions in the modeling set-up which lead to the result that the largest firm prefers a lower level of interconnection quality compared to its smaller rivals. Below it is discussed whether these assumptions are relevant for the market for Internet backbone services. Market entry conditions First, consider the assumption of a fixed number of firms in the market. This assumption does not correspond well to the thousands of active ISPs observable in reality. If at all, then this assumption may apply to the market for Tier-1 ISP services in which only 5 to 10 ISPs are active. But whether this market has structural barriers 111 Crémer, Rey and Tirole (ibid., 458) a gue that the dominant firm n limit the capacity of the interface with the non-targeted network to such an extent that the capacity is only sufficient to provide good quality interconnection for the traf ic of the non-targeted network but would resul in very bad interconnection quality if the traffic should grow to encompass also the traffic of the targeted network. 126 to entry, which justify the assumption of a fixed number of firms, is what is trying to be proved. To start with this assumption distorts the analysis of the effects of network externalities on competition in this market. The equilibrium results of the model by Crémer, Rey and Tirole change dramatically when the number of firms in the market is endogenized (Malueg and Schwartz: 2006). Consumers do not necessarily choose the firm with the initially larger installed base. When this firm chooses not to be compatible with it’s smaller rivals,112 and when smaller rivals in sum have a minimum initial market share and choose to remain compatible among themselves, then, for a large set of parameter values, new consumers will sign on to this network of smaller compatible firms in the expectation that in a dynamic market setting this network will eventually incorporate more contacts than the single-firm network of the initial market leader.113 If payments for interconnection were introduced, the parameter values for which the initially larger firm would choose autarky would be even more limited because smaller firms could share their gain from increased connectivity by offering payments to the larger firm. The assumption that smaller rivals will remain compatible amongst one another, in the sense that they remain interconnected, and will have a significant network reach through this interconnection of their networks is very realistic for the Internet backbone services market. According to Crémer, Rey and Tirole, a strategy of secondary peering among lower-level ISPs can be of only limited influence on the backbone market, as the “setting up of new connectivity structure would involve a lag, since one must build bilateral interconnections and test new routing procedures” (Crémer, Rey and Tirole, 2000: 446). In reality, new interconnection agreements are, however, entered into easily. The presence of many ISPs at Internet exchange points and the availability of standardized contracts facilitate interconnection agreements. Lower level ISPs can enter into secondary-peering relations simply by opening their networks to one another’s traffic, by engaging in BGP sessions. The only traffic that secondary ISPs will never be able to get rid off by secondary peering and transit agreements among lower network levels is the traffic which is destined to or originating from direct subscribers of Tier-1 ISP, when these ISPs have no peering contracts with lower level networks and when the customers are not multi-homed so that they cannot be reached by alternative paths. Interconnection 112 The targeted degradation scenario is not considered by Malueg and Schwartz. In a related working paper (Malueg and Schwartz, 2002: 37) the authors argue that the parameter values that make targeted degradation profitable to the dominant firm imply unrealistic values for price relative to marginal cost and the consumer surplus of the median subscriber. 113 Even when the dominant network’s installed customer base is larger than the combined installed customer bases of it’s rivals, there are parameter regions in which the rivals will be more successful in adding new customers to their networks (Malueg and Schwartz, 2006: 9). This is due to the customers’ expectations of market evolvement in dynamic market settings, in which networks are expected to have a high growth potential. This conclusion is comparable to the results by Economides (1996) for a monopolist that prefers inviting market entry. 126 to entry, which justify the assumption of a fixed number of firms, is what is trying to be proved. To start with this assumption distorts the analysis of the effects of network externalities on competition in this market. The equilibrium results of the model by Crémer, Rey and Tirole change dramatically when the number of firms in the market is endogenized (Malueg and Schwartz: 2006). Consumers do not necessarily choose the firm with the initially larger installed base. When this firm chooses not to be compatible with it’s smaller rivals,112 and when smaller rivals in sum have a minimum initial market share and choose to remain compatible among themselves, then, for a large set of parameter values, new consumers will sign on to this network of smaller compatible firms in the expectation that in a dynamic market setting this network will eventually incorporate more contacts than the single-firm network of the initial market leader.113 If payments for interconnection were introduced, the parameter values for which the initially larger firm would choose autarky would be even more limited because smaller firms could share their gain from increased connectivity by offering payments to the larger firm. The assumption that smaller rivals will remain compatible amongst one another, in the sense that they remain interconnected, and will have a significant network reach through this interconnection of their networks is very realistic for the Internet backbone services market. According to Crémer, Rey and Tirole, a strategy of secondary peering among lower-level ISPs can be of only limited influence on the backbone market, as the “setting up of new connectivity structure would involve a lag, since one must build bilateral interconnections and test new routing procedures” (Crémer, Rey and Tirole, 2000: 446). In reality, new interconnection agreements are, however, entered into easily. The presence of many ISPs at Internet exchange points and the availability of standardized contracts facilitate interconnection agreements. Lower level ISPs can enter into secondary-peering relations simply by opening their networks to one another’s traffic, by engaging in BGP sessions. The only traffic that secondary ISPs will never be able to get rid off by secondary peering and transit agreements among lower network levels is the traffic which is destined to or originating from direct subscribers of Tier-1 ISP, when these ISPs have no peering contracts with lower level networks and when the customers are not multi-homed so that they cannot be reached by alternative paths. Interconnection 112 The targ ted d gradation scenario is not c nsidered by Malueg and Schw rtz. In r lated working paper (M lueg and Schwartz, 2002: 37) the authors argue that the parameter values that mak targ ted degradation p ofitable to the dominant firm imply unrealistic values for price relative to marginal cost and the consumer su plus of the med an subscriber. 113 Even when the dominant network’s in talled custome base is larger than the combined installe custom r bases of it’s rivals, there are parameter regions in which the rivals will be more successful in dding new customers to their networks (M ueg and S hwartz, 2006: 9). This is du to the customers’ expectations f market evolvement in dynamic market settings, in which networks are expected to have a high growth potential. This conclusion is comparable to the results by Economides (1996) for a monopolist that prefers inviting market entry. 127 agreements are historically grown and peering was more common at the beginnings of the Internet. Some secondary ISPs therefore continue to have peering with a limited number of the top-level ISPs. They can use these peering relationships to offer transit services to other lower-level ISPs. Considering also that many subscribers of Internet services are multi-homed (i.e. subscribe to several networks) it becomes clear that the Internet reach provided to the customers of the lower-level ISPs can be increased significantly by coordination on the lower hierarchy levels. Elixmann and Scanlan (2002: 111) argue that the proportion of traffic in the Internet which can theoretically be routed via secondary peering has increased. Lower-level ISPs may be less dependent on Tier-1 ISPs than suggested in some of the literature. Product differentiation Secondly, consider the assumption made by Crémer, Rey and Tirole that customers do not have individually differing preferences for technology characteristics of the network they subscribe to. This assumption does not correspond well to the reality of a large degree of product differentiation observable among ISPs. On the Internet backbone services market, ISPs offer their services to other ISPs, to web-hosting services, to large business users and to private end-users. They offer different service levels according to their customers’ needs and they offer their services at diverse locations, again according to their customers’ needs. An ISP that would hope to make the market tip in its favor would have to cater to all customers in the market. This may not be the most profitable market strategy in a world of customer heterogeneity. ISPs that focus on particular customer groups have comparative advantages in supplying the types of services that these customers prefer. In this case, the proper theoretical reference model may be that ISPs are supplying components of systems rather than competing systems. In such markets, compatible products (as, for instance, interconnected networks) cater to the needs of particular customers. Competition between the products is not as strong as in a market of competing systems because the possibility to make profits is often increased by compatibility (see Economides, 1989 and Einhorn, 1992). When product differentiation is introduced into the model set-up it can be shown that in any shared market equilibrium both firms profit from a higher interconnection quality because competition in other product dimensions becomes less aggressive when firms offer the same positive network externality effect to their customers (Foros and Hansen, 2001).114 General analysis on the compatibility incentives of providers of differentiated network goods come to comparable results (Doganoglu and Wright, 2006). Switching costs There are other critical assumptions in the model by Crémer, Rey and Tirole which do not correspond to the characteristics of the Internet backbone services market. 114 In this model there is also no installed customer base. This fact of course also has an important impact on the results of the model. This aspect is in the focus of a model structure by Economides (2005) which is discussed below. 128 Firstly, consider the assumption that installed bases are locked-in. In reality, switching ISPs is not difficult for end-users or ISPs. Only the cancellation period of their contract may delay the reaction for some weeks. Larger customers such as firms and ISPs are often multi-homed, that is, they connect to more than one ISP at any given time. This is important for the ISP to be able to guarantee its contractual service level vis-à-vis its customers. It is also a signal that traffic can be diverted quickly from one ISP to another without large transaction costs involved. The fact that switching is relatively easy increases the competition between Internet backbone service providers. When the assumption of a locked-in customer base is relaxed, it can be shown that the initially dominant network has an incentive to keep up a high quality of interconnection (Economides, 2005: Appendix). A degradation of interconnection quality with one of the smaller rivals would lead to a loss of universal connectivity that would result in a severe demand response by the installed customer base and therefore to revenue and profit loss. Crémer, Rey and Tirole argue that multi-homing may even increase the incentive to degrade interconnection quality for a large ISP. They assume that the demand reduction effect, which results from a decrease in interconnection quality, is relatively less pronounced when more customers are multi-homed and can reach users of other networks over alternative connections (Crémer, Rey and Tirole, 2000: 465). What they do not consider is that if customers use their alternative interconnection agreements, then they can shift most or all of their demand away from the network with reduced interconnection quality. A significant demand reduction can therefore be the result of customers having alternative providers of interconnectivity. Demand In the model by Crémer, Rey and Tirole network externalities are depicted by a demand function which is linear in the number of customers which can be reached via the network. In reality, the number of customers attached to any ISP is intransparent to users. There are thousands of active ISPs and it is nearly impossible for an end-user to know roughly how many customers are connected to these ISPs, let alone know which particular user can be reached by which particular ISP. Because the connections and interconnections in the Internet are so intransparent, Economides (2005) argues that it is not the number of customers reachable via a network that users are interested in, but rather whether a network offers universal connectivity or not. A high network coverage, once it is below universal connectivity, is a significant quality reduction from the point of view of the end-user because Internet users cannot know which single connections they will value highly in the future and which ISP is likely to include these connections in a world of limited connectivity. Economides (2005: 400) assumes that “…no network, however large, can afford not to offer universal connectivity.” It is true that some sites receive far more traffic than other sites and that it is extremely important for an ISP to provide good quality interconnection to these highly frequented destinations (i.e. web servers with popular content). However, an ISP 129 cannot use sole access to popular content to act as a counterbalance for reduced overall connectivity. Content providers would not accept a reduced reachability. If the dominant ISP had exclusive connectivity to a popular website, then the owner of this website would demand to be reachable universally. If the ISP degrades interconnection with some parts of the Internet, then the owner of the website will move its content to ISPs which continue to have good interconnectivity with the entire Internet. 7.4 Collusion on the Tier-1 level Only Tier-1 ISPs can guarantee universal connectivity without relying on a transit offer. The preceding section showed that one Tier-1 alone cannot successfully refuse interconnection with other ISPs or raise interconnection prices in the hopes of ousting competitors from the market. The transit offers of Tier-1 ISPs are perfect substitutes. Absent any collusive practices there is intense competition in this market. This fact provides the Tier-1 ISPs with a motive to collude on the market for transit services. If all Tier-1 ISPs acted simultaneously in increasing prices for transit services, then lower level ISPs would have no alternative transit provider from whom to buy universal connectivity services. And no new provider of universal connectivity could enter the market as long as the Tier-1 ISPs would successfully foreclose this market by not entering into any new peering agreements. The question analyzed in the present section is whether Tier-1 ISPs can organize a stable collusion in the wholesale market in order to collectively raise the price of transit services? There is a literature on two-way access in telecommunications markets which analyzes whether cooperation on the wholesale level can help enforce collusion on the retail level.115 A two-way access scenario is given when customers connect to only one network, such that the two networks reciprocally need access to each other’s customers on the wholesale level. Termination in this scenario is comparable to a monopolistic bottleneck. The application of this literature has mostly been to voice telephony markets, for instance, mobile telephony or reciprocal international termination. Considering that ISPs have a termination monopoly whenever customers exclusively connect to their network only, the models may, however, also be applicable to the market for Internet backbone services. If a large fraction of endusers are connected to only one network, then ISPs may have the possibility to collude on the retail market. The assumptions that are necessary for successful collusion in a market with reciprocal termination are: • There is no free market entry. • There are no capacity limitations. • Every customer connects to only one network. 115 The seminal articles in this research are Laffont, Rey and Tirole (1998a and 1998b).