115 7 Competitive effects of network externalities in the context of the Internet The focus of the present chapter is on the strong network externalities associated with Internet service provision and their effect on competition in Internet backbone services. In chapter 5 it was shown that there are no monopolistic bottleneck network elements in Internet backbone services. From the point of view of the theory of monopolistic bottlenecks ex-ante sector-specific regulation cannot be justified on the logical layer of Internet service provision. If no monopolistic bottleneck network element can be found, competition authorities can analyze a particular market to ascertain whether general competition policy intervention into a market is deemed necessary. The market for Internet service provision was analyzed by competition authorities in the context of the merger proceedings regarding the merger of Worldcom and MCI in 1998 and the failed merger of MCI/Worldcom and Sprint Corporation in 2000. The focus in these analyses was on the strong network externalities in the market for Internet service provision. The general discussion of network externalities in chapter 6 shows that in many market constellations firms have incentives to voluntarily cooperate in order to internalize positive network externalities. These incentives are especially strong when customers have heterogeneous preferences, when products are differentiated and when the possibility for compensation between the firms is given. With respect to the network externalities in the market for Internet service provision, competition authorities doubted that competitive forces in the market were strong enough to cope with a merger of two of the four largest firms in the market for Internet backbone services. The analyses conducted in the context of the merger proceedings in the market for Internet backbone services are reviewed in the present chapter in order to determine whether the disaggregated analysis of the market for Internet service provision captured all relevant aspects of competition in Internet service provision which should be included in a comprehensive examination of the question whether sector-specific regulation is justified in the market for Internet service provision. Of course, the focus of the competition policy analyses reviewed below was not to localize market power which would justify regulation. However, the findings with respect to the functioning of competition in Internet service provision should be accounted for in the present analysis. The chapter is structured as follows: Section 7.1 discusses the network externalities in the applications layer of Internet service provision and how they relate to the logical layer of Internet service provision. Section 7.2 reviews the terms of interconnection between ISPs in competitive interconnection markets. Section 7.3 reviews the literature on interconnection incentives of ISPs with a focus on the singledominance case. Section 7.4 analyzes whether Tier-1 ISPs as a group could success- 116 fully form a stable collusion on the market for transit services and thereby collectively discriminate ISPs on lower hierarchy levels (collective dominance). Section 7.5 concludes the chapter. 7.1 Network effects in Internet service provision The Internet, as a classical communications network, belongs to the class of goods which exhibit positive external benefits in consumption. Direct external effects are due to the fact that the utility of belonging to the Internet community is directly related to the number of other people and services that can be reached via the Internet. Indirect network effects result from the fact that the more people use Internet services, the more applications and complementary products are offered to Internet users. The utility derived from the consumption of any network good can be decomposed into a so-called network effect, resulting from the number of people reachable via the network, and a so-called technology effect, resulting from the technological characteristics of the network the user is connected to (Blankart and Knieps, 1992: 80). In the context of Internet service provision the network effect can be expected to dominate the technology effect because users are more likely to give up benefits from a preferred technology for a wider reach in the Internet. One way of maximizing the benefits from the network effect in Internet services is to have only one network supply its services to all users. This would, however, imply that consumers can derive no benefits from competition among ISPs over price, product or service quality. As an alternative to a single large network, network interconnection among otherwise independent network operators can allow that users enjoy the positive network externalities associated with a single network while benefiting from product diversity in product dimensions other than network size. Indeed, the principal attraction of the Internet is that because of interconnection among ISPs anyone connected to the Internet is reachable by all other users of the public Internet, irrespective of the home ISPs these users subscribe to. Internet users expect this universal connectivity from their ISP, that is, the ability to reach all destinations reachable on the public Internet. For universal connectivity all networks need to be either directly or indirectly connected to one another. The strong network effect experienced on the retail level of Internet services provision therefore translates into a demand for Internet interconnection by ISPs on the logical layer of Internet service provision. Still, an ISP’s incentives for interconnection may be contradictory, when on the one hand an ISP wants to offer universal connectivity to its customers and therefore will seek to interconnect with rival networks, but on the other hand it could try to gain a competitive advantage by refusing to interconnect with some ISPs, thereby keeping them out of the market and trying to lure the customers of these ISPs to its own network instead. 117 7.2 Terms of interconnection among IP-based networks in a competitive market environment The interconnection of networks has three aspects. Firstly, a logical interconnection of the networks needs to define which services are to function across the network boundaries and at which quality. Secondly, a physical interconnection between the network infrastructures needs to be established. Lastly, the ISPs need to negotiate how the costs of the physical interconnection and the costs for the traffic transmission via this interconnection ought to be split. The advantage of the TCP/IP standard is that two IP-based networks can agree to use the TCP/IP protocol and thereby define much of what the logical interconnection parameters will be. ISPs can negotiate further quality of service parameters which they want to guarantee across network boundaries. Advanced services, such as realtime Voice over Internet Protocol (VoIP) capabilities or television over Internet Protocol services can, for instance, be offered only to users within one and the same network by running additional protocols on top of the standard TCP/IP protocols.102 They can however, also be offered across network boundaries, if the ISPs agree to guarantee the required quality parameters. Negotiations over physical interconnection as well as the financial terms of network interconnection need to address the following questions: 1) where to establish the location of the interconnection, 2) how to cover the costs of the network infrastructure which physically connects the two networks and 3) how the two networks ought to split the costs for traffic transmission to and from the other’s network. The following subsections present the typical financial agreements for Internet transport services today. 7.2.1 Costing and pricing of Internet transport services Early interconnection of IP-based networks in the NSFNET era functioned basically without monetary compensation between the connecting parties. The rationale may have been that traffic flows could be expected to be roughly symmetrical. More importantly, however, the funding for the network infrastructure at this time was in most cases provided by the government. Network administrators therefore considered the effort to install complex traffic metering dispensable. This situation changed fundamentally, when the National Science Foundation reduced funding and networks had to become self-supporting. It was especially important for the newly emerging commercial ISPs to operate cost-effectively. Network costs needed to be recovered according to some cost-causation principle. It is no coincidence that inter- 102 See, for instance, Buccirossi et al. (2005). See also section 9.3 in which the implications of the more recently developed technologies for quality of service differentiation among networks are discussed. According to Marcus (2006b: 34) these technologies are already widely deployed for controlling the quality of service within networks.