167 10 Summary The objective of the present thesis was to provide an economic analysis of the supply-side and demand-side characteristics of the market for Internet service provision with a view to answering the question of whether there is a role for sectorspecific regulation in this market. Policy concerns regarding Internet service provision stem from the typical network-industry characteristics of this market. Economies of scale and scope on the cost side as well as network effects on the demand side give reason to suspect market failure. Much of the literature on competition in Internet service provision deals with the question whether large Tier-1 ISPs have less incentive for network interconnection than their smaller rivals. More recently, the question whether providers of Internet transport services have an incentive to discriminate against independent Internet applications and content providers has gained in importance (the network neutrality concern). The debate over these competition concerns has had actual consequences for mergers of firms active in Internet service provision, which have either been blocked entirely or were granted only subject to obligations. Because understanding the specifics of the Internet is crucial for an analysis of the competitive forces that are present in the market, the analysis started out by describing the basic characteristics of Internet service provision and by defining the terms that are used throughout the thesis, to refer to different parts of the market for Internet service provision. Within Internet service provision the analysis distinguishes between Internet core services and Internet periphery services. Internet application services and Internet transport services belong to the Internet core because they have viable functions only within the Internet. The physical network layer and the content layer of Internet service provision are characterized by the fact that they have viable functions apart from their role in Internet service provision. They therefore belong to the Internet periphery. The elements of Internet service provision were then categorized according to a general layered model of communications networks. This model was used as a starting point for identifying the need for sector-specific regulation in Internet services. The differentiation between Internet core and Internet periphery allows one to make a distinction between market power problems located in the Internet core and market power problems located in the Internet periphery. As a further basis for the subsequent analysis of the possible need for sectorspecific regulation in Internet service provision the thesis gives an overview of the historical and technical background of the Internet. This description gives the reader detailed information on Internet technology and Internet interconnection. The history of the Internet was reviewed with a special focus on the development of network interconnection between ISPs. The three-tier hierarchy of the Internet was explained. 168 Furthermore, both the technical and the commercial terms for Internet interconnection were explained. The introductory chapters to the analysis showed that all ISPs can be considered essentially equal in the sense that all are active on the logical layer of Internet services provision, and all require either cooperation with firms active on the other layers of Internet services provision or need to vertically integrate into these layers themselves in order to be able to offer services to end-users. Upon closer examination it was shown that the services offered on the logical layer of Internet services provision are different depending on whether one is considering a Tier-1 ISP or a lowerlevel ISP. Only Tier-1 ISPs have the ability to offer universal connectivity to endusers on their own account, without resorting to a transit provider. All other ISPs need to purchase transit services from at least one Tier-1 ISP, either directly or indirectly through intermediate ISPs so as to be able to guarantee universal connectivity in their network. Universal connectivity is a quality characteristic which has become essential to Internet users. One relevant question for the subsequent competition analysis was therefore whether the fact that only Tier-1 ISPs can offer universal connectivity on their own account lends market power to these ISPs. Since competition policy and sector-specific regulation have many limitations, the analysis of the competition policy concerns in Internet service provision started from the assumption that market failures must be severe to justify government intervention. Generally, the appraisal of competition in network industries requires a reference model for competition, which allows for the typical characteristics of network industries without automatically equating these with market power. The theory of monopolistic bottlenecks was introduced as a reference model which was designed explicitly for localizing stable market power in network industries. According to this theory, only monopolistic bottleneck network areas, defined as market areas with both natural monopoly characteristics as well as sunk costs, lend stable network-specific market power to the owner of the bottleneck element. Ex-ante sector-specific regulation of network sectors can only be justified, when such monopolistic bottlenecks are found. The disaggregated regulatory approach was introduced as a regulatory framework which uses the theory of monopolistic bottlenecks to localize monopolistic bottlenecks in network industries. This approach was applied to the layered model of Internet service provision. The disaggregated analysis of the network elements in the value chain of Internet service provision lead to the conclusion that, at least in some geographic regions, parts of the local communications infrastructure located on the physical layer of Internet service provision must be considered monopolistic bottleneck network elements. There are no monopolistic bottleneck network elements in long-distance network capacity or on the logical layer of Internet service provision, which are at the center of the discussion on potential Internet regulation. Network externalities are often cited as the second possible cause for market failure in the market for Internet service provision. To establish a general understanding of the effects of network externalities on competition in network industries, the analysis first reviewed relevant literature on how market processes are affected by 169 network externalities. It was shown that market processes very often generate novel ways for coordinating users and firms – especially in markets with heterogeneous consumers and product differentiation. In many realistic market environments the spontaneous market order has a comparative advantage in internalizing network externalities as compared to administered policy. It was therefore argued that to support government intervention as a means of internalizing network externalities requires strong evidence that the many ways by which market participants can solve the information problem are not functional under particular circumstances, as for instance in the case of an uncontestable natural monopoly. With respect to the network externalities in the market for Internet service provision, competition authorities feared that a merger between two of the large Tier-1 ISPs would put the resulting firm in a position in which it could charge discriminatory rates for its transit services. The competition policy analyses that were conducted in the context of the merger proceedings in the market for Internet backbone services were reviewed 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. The review of the competition policy analyses showed that the fear that a single Tier-1 ISP could be able to abuse a dominant market position in a transit agreement with lower-level ISPs cannot be substantiated when the strong competitive forces in the market for top-tier Internet backbone services are taken into account. Tier-1 ISPs compete with sophisticated product differentiation tactics. Furthermore, customers frequently multi-home and can relatively conveniently switch their home network. As a result, Tier-1 ISPs cannot benefit from refusing to interconnect respectively from deteriorating interconnection quality with lower-level networks. The analysis also examined whether a collusion between Tier-1 ISPs, to collectively raise prices in the transit market, is likely to be stable. It was shown that the prerequisites for a stable collusion are not fulfilled in the market for top-tier Internet backbone services, most importantly because the assumption of a termination monopoly is not fulfilled. A necessary prerequisite for the supposed effective competition in Internet transport markets is non-discriminatory access regulation of any remaining monopolistic bottleneck network areas in the upstream markets on the physical layer of Internet service provision. A disaggregated regulation of the monopolistic bottleneck elements would need to adhere to the following principles: (1) Market intervention by regulation should be applied to monopolistic bottleneck network areas only. (2) Regulation should be symmetrical in the sense that all market participants should be treated equally when applicable (for instance, incumbents and new entrants should all contribute in equal measure to the policy goal of achieving universal service). (3) Accounting separation should be required of the regulated firm together with information regarding quality of service vis-à-vis own divisions and third parties. (4) Price-cap regulation should be employed in the regulation of the access charge for monopolistic bottlenecks. 170 The implementation of the disaggregated regulatory framework to the monopolistic bottlenecks in local communications infrastructure begins with a clear demarcation of the monopolistic bottleneck network areas. From the viewpoint of a policy maker interested in ensuring competition in Internet service provision it was shown that the monopolistic bottleneck in traditional PSTN telecommunications networks comprises the copper cable connecting the customer premises with the incumbent’s MDF-site in those regions in which infrastructure-platform competition has not emerged and is not likely to emerge. Access to the bare copper cable allows entrants to differentiate their end-user services from the services of the incumbent also in technological characteristics. In rural areas entrants will furthermore often require access to co-location space to set up their network equipment. In an NGN environment the extent of the monopolistic bottleneck network area decreases. Whenever there is no infrastructure-platform competition from cable-based or wireless operators in the market for high-speed Internet access, it was argued, that nondiscriminatory access to the ducts and the ductworks of the incumbent’s local access network can put competitors in a position to build out fiber cables closer to the enduser premises at costs equal to those of the incumbent. Competitors will further require non-discriminatory access to the remaining wiring between the end-user’s outlet and the network point to which the fiber has been built out. The evaluation of current regulatory practices in the United States showed that U.S. telecommunications legislation, in theory, comes close to the regulation guidelines derived in the normative analysis. The practical implementation of regulation in the U.S., however, does not always conform to the legal standard. Most importantly, there is no open-access regulation for broadband communications infrastructures, even when no second communications platform exists or is likely to be built. The lack of broadband-access regulation in the U.S. is at least partly responsible for the fear of network neutrality activists, that vertically integrated incumbents may have an incentive to discriminate competitors in Internet service provision markets. The FCC should use the instruments provided by the Telecommunications Act of 1996 to regulate the monopolistic bottleneck network elements that competitors need to offer competitive high-speed Internet access services. In Europe, on the other hand, both the regulation legislation as well as its practical implementation show a strong tendency to overregulation. The European regulatory framework for communications markets regulation lacks a clearly defined limiting criterion for the implementation of ex-ante sector-specific regulation. The upcoming reform of the EU regulatory framework for electronic communications should be used to specify a clear criterion by which monopolistic bottlenecks can be identified. Regulation should be restricted to these network areas and lifted in all other cases. Current regulatory practice in telecommunications markets differentiates between wholesale products for voice telecommunications services and wholesale inputs for Internet service provision. Once the convergence of telecommunications, Internet, and broadcasting networks is completed, this differentiation will become obsolete. In the near future, NGNs will serve as the common physical layer for all communi- 171 cations services. The deployment of NGN infrastructure is well underway. European incumbents, for instance, expect to migrate to all-IP based core networks as early as 2010 (KPN, Netherlands and T-Com, Germany)165 and 2011 (BT, UK).166 The migration to all-IP networks in local access markets will be realized in a second step. Regulators and industry representatives are now discussing whether new NGN regulations will soon have to replace existing PSTN regulations in order to ensure a competitive market for electronic communications services.167 It is reasonable to assume that the interconnection conventions of today’s Internet will be carried over into the NGN world. Carriers will no longer charge for origination and termination of single connections as is common in today’s PSTN networks. They will rather implement a system of flat-rate end-user charges in combination with transit payments and peering agreements on the wholesale level. Offers for different QoS transport services for content providers may complement these charges and provide further revenue to network owners. NGN interconnection will soon need to be incorporated into the regulatory framework for electronic communications markets. The present thesis developed the extent of the monopolistic bottleneck in NGN networks and showed to which network elements open-access regulation needs to be applied. Before NGN migration is complete, regulatory authorities must decide on the time frame during which incumbents should be obliged to offer current wholesale PSTN products in parallel to new NGN wholesale products. As was shown in the analysis, the extent of the monopolistic bottleneck in the local communications infrastructure decreases in an NGN environment. Local switching facilities at the MDF site of the incumbent can, for instance, be considered part of the monopolistic bottleneck in PSTN networks but not in NGN networks. Incumbents should not be required to continue operations of a PSTN network architecture once their NGN network architecture is completed. On the other hand, competitors should, at the minimum, be guaranteed the PSTN wholesale products for the contractual terms laid down in current interconnection agreements. Since competitors need to adapt their network architecture to the new regulatory environment, regulators may even prolong current regulations for some additional time. Independent of the time horizon decided on for the continuation of PSTN regulations, it is important that the decision for the migration of regulation be communicated early and clearly, and that incumbents and new entrants alike can trust in the stability of this decision. 165 See:http://www.portel.de/nc/nachricht/artikel/wik-vdsl-neuer-markt-oder-neueanschlusstechnologie-im-ngn/ ; site last visited on Feb. 16, 2008. 166 See: http://www.btplc.com/21CN/Theroadto21CN/Keymilestones/Keymilestones.htm; site last visited on Feb. 16, 2008. 167 See, for instance, Berg et al. (2006) and ERG (2007).