TOWARDS AN ECONOMIC FRAMEWORK FOR NETWORK NEUTRALITY REGULATION

Electronic copy available at: http://ssrn.com/abstract=812991

Towards an Economic Framework for Network Neutrality Regulation

Dr.-Ing. Barbara van Schewick, Ass. iur.

Published Version This paper has been published by the Journal on Telecommunications and High Technology Law in 2007. Please cite as ‘Barbara van Schewick, Towards an Economic Framework for Network Neutrality Regulation, 5 J. ON TELECOMM. & HIGH TECH. L. 329’. This paper can be downloaded without charge from the Social Science Research Network Electronic Paper Collection: http://ssrn.com/abstract=812991

Electronic copy available at: http://ssrn.com/abstract=812991

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

329

TOWARDS AN ECONOMIC FRAMEWORK FOR NETWORK NEUTRALITY REGULATION

DR.-ING. BARBARA VAN SCHEWICK, ASS. IUR.*

Network neutrality rules forbid network operators from excluding or discriminating against third-party applications. This analysis shows that calls for network neutrality regulation are justified: absent net- work neutrality regulation, network providers will likely discriminate against or exclude independent producers of applications, content, or portals from their networks. This threat reduces the amount of inno- vation in applications, content and portals at significant costs to so- ciety. While network neutrality rules remove this threat, they are not without costs. Due to the potentially enormous benefits of applica- tion-level innovation for economic growth, however, increasing the amount of application-level innovation through network neutrality regulation is more important than the costs associated with it. This paper also highlights important limitations of the “one monopoly rent” argument, demonstrating previously unidentified exceptions that may be quite common in the Internet context, showing how ex- clusion may be a profitable strategy even if the excluding actor does not manage to drive its competitors from the complementary market, and proving that competition in the primary market may be insuffi- cient to remove the ability and incentive to engage in exclusionary conduct.

* Non-Residential Fellow, Center for Internet and Society, Stanford Law School, CA, USA and Senior Researcher, Telecommunication Networks Group, Department of Electrical Engineering and Computer Science, Technical University Berlin, Germany. E-Mail: schewick@stanford.edu / schewick@tkn.tu-berlin.de.

Thanks to Pio Baake, Yochai Benkler, Marjory Blumenthal, David Clark, Joseph Farrell, Gerald Faulhaber, David Isenberg, Bill Lehr, Lawrence Lessig, Doug Lichtman, Robert Pep- per, Arnold Picot, David Reed, Tim Wu, Christopher Yoo and participants of the 33rd Re- search Conference on Communication, Information and Internet Policy (TPRC 2005), the Berkman Center for Internet and Society Luncheon Series, Harvard Law School, and the MIT Computer Science and Artificial Intelligence Laboratory Speaker Series, Massachusetts Insti- tute of Technology, for comments on an earlier version of this paper and for discussions.

Parts of this paper are based on Barbara van Schewick, Architecture and Innovation: The Role of the End-to-End Arguments in the Original Internet, Chapter 9 and Chapter 11 (Ph.D. dissertation, Technical University Berlin 2005, MIT Press forthcoming 2008), for which finan- cial support of the German National Academic Foundation (“Studienstiftung des Deutschen Volkes”) and the Gottlieb Daimler- and Karl Benz-Foundation is gratefully acknowledged.

0502_006_VANSCHEWICK 2/19/2007 10:28:10 AM

330 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

INTRODUCTION …………………………………………………………………………. 331 I. THREAT OF DISCRIMINATION …………………………………………………… 336 A. Stylized Model ……………………………………………………………….. 337 B. Network Provider is Monopolist in the Market for Internet Services ……………………………………………………………… 340 1. No General Incentive to Discriminate…………………………. 340 2. New Exceptions……………………………………………………….. 342 2.1. Complementary Product Source of Outside Revenue……. 342 2.2. Monopolist’s Complementary Product Source of Outside Revenue………………………………………………………………….. 345 2.3. Monopoly Preservation in the Complementary Market…. 347 3. Relevance of Known Exceptions ……………………………….. 353 3.1. Primary Good not Essential ………………………………………. 353 3.2. Monopoly Preservation in the Primary Market …………….. 357 4. Profitability of Discrimination without Monopolization… 364 4.1. More Sales at Market Prices………………………………………. 365 4.2. More Outside Revenue……………………………………………… 367 4.3. Monopoly Preservation in the Complementary Market …. 368 C. Network Provider Faces Competition in the Market for Internet Services ……………………………………………………………… 368 1. Ability to Exclude ……………………………………………………. 371 2. Benefits of Exclusion ……………………………………………….. 372 2.1. More Sales at Market Prices………………………………………. 373 2.2. More Outside Revenue……………………………………………… 373 2.3. Monopoly Preservation in the Complementary Market …. 374 2.4. Preserving Competitive Position in the Primary Market… 374 3. Costs of Exclusion……………………………………………………. 375 D. Conclusion……………………………………………………………………… 377 II. IMPACT ON APPLICATION-LEVEL INNOVATION …………………………. 378 A. Incentives of Independent Producers of Complementary Products …………………………………………………. 378 B. Incentives of Network Providers ……………………………………….. 380 III. IMPACT ON SOCIAL WELFARE………………………………………………… 382 A. Benefits………………………………………………………………………….. 383 B. Costs ……………………………………………………………………………… 386 1. Impact on Incentives at the Network Level………………….. 386 2. Costs of Regulation ………………………………………………….. 387 C. Trade-Off ……………………………………………………………………….. 387 1. Application-Level Innovation vs. Innovation at the Network Level…………………………………………………………. 387 2. Application-Level Innovation vs. Deployment of Network Infrastructure ……………………………………………… 388 CONCLUSION…………………………………………………………………………….. 390

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 331

INTRODUCTION

Over the past years, the merits of network neutrality regulation have become a hot topic in telecommunications policy debates. Repeatedly, proponents of network neutrality regulation have asked the Federal Communications Commission to impose rules on the operators of broad- band access networks that forbid network operators to discriminate against third-party applications, content or portals (“independent applica- tions”) and to exclude them from their network.1 Congress is currently considering proposals to introduce network neutrality legislation;2 the House of Representatives and the Senate held hearings on the subject.3

1. See, e.g., Ex parte Submission of Tim Wu and Lawrence Lessig to the Declaratory Ruling & Notice of Proposed Rulemaking in Inquiry Concerning High-Speed Access to the Internet, CS Dkt. No. 02-52 (Aug. 22, 2003), available at http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6514683885 [hereinafter Wu & Lessig, Ex parte]; Comments of the High Tech Broadband Coalition, to the Declaratory Ruling & Notice of Proposed Rulemaking in Inquiry Concerning High-Speed Ac- cess to the Internet, CS Dkt. No. 02-52 (June 17, 2002), available at http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6513198353; Ex parte Submission of the Coalition of Broadband Users and Innovators to the Declaratory Ruling & Notice of Proposed Rulemaking in Inquiry Concerning High-Speed Access to the Internet, CS Dkt. No. 02-52 (July 17, 2003), available at http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6514286197. For proponents of nondiscrimination rules in the scientific arena, see, for example, LAWRENCE LESSIG, THE FUTURE OF IDEAS 248-49 (2001) [hereinafter LESSIG, FUTURE OF IDEAS]; Philip J. Weiser, Toward a Next Generation Regulatory Strategy, 35 LOY. U. CHI. L.J. 41 (2003); Wu & Lessig, Ex parte, supra; Tim Wu, Network Neutrality and Broadband Discrimination, 2 J. ON TELECOMM. & HIGH TECH. L. 141 (2003) [hereinafter Wu, Network Neutrality]; Tim Wu, The Broadband Debate: A User’s Guide, 3 J. ON TELECOMM. & HIGH TECH. L. 69 (2004) [hereinafter Wu, Broadband Debate]; Brett M. Frischmann & Barbara van Schewick, Yoo’s Frame and What it Ignores: Network Neutrality and the Economics of an Information Super- highway, 47 JURIMETRICS (forthcoming 2007); Bill D. Herman, Opening Bottlenecks: On Be- half of Mandated Network Neutrality, 59 FED. COMM. L. J. (forthcoming 2007), available at http://ssrn.com/abstract=902071; Robert D. Atkinson & Philip J. Weiser, A Third Way on Net- work Neutrality, THE NEW ATLANTIS, Summer 2006, at 47, available at http://www.thenewatlantis.com/archive/13/atkinsonweiser.htm (last visited November 23, 2006); Susan P. Crawford, Network Rules, LAW & CONTEMP. PROBS. (forthcoming 2007), available at http://ssrn.com/abstract=885583. 2. For an overview of the different proposals and their history, see Declan McCullagh, Republicans Defeat Net Neutrality Proposal, CNET NEWS.COM, Apr. 5, 2006, http://news.com.com/2100-1028_3-6058223.html. For an overview of government actions and statements of officials concerning network neutrality, see John Windhausen, Public Knowl- edge, Good Fences Make Bad Broadband. A Public Knowledge White Paper, http://static.publicknowledge.org/pdf/pk-net-neutrality-whitep-20060206.pdf (Feb. 6, 2006) [hereinafter Public Knowledge White Paper], at 13-16. 3. Net Neutrality: Hearing Before the S. Comm. on Commerce, Science, and Transpor- tation, 109th Cong. (2006), available at http://frwebgate.access.gpo.gov/cgi- bin/getdoc.cgi?dbname=109_senate_hearings&docid=f:30115.pdf [hereinafter Senate Hear- ing]; Internet Protocol and Broadband Services Legislation: Hearing Before the Subcomm. on Telecommunications and the Internet of the H. Comm. on Energy and Commerce, 109th Cong. (2005), http://frwebgate.access.gpo.gov/cgi- bin/getdoc.cgi?dbname=109_house_hearings&docid=f:26998.pdf.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

332 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

Network neutrality proposals are based on the concern that in the absence of such regulation, network operators may discriminate against independent applications and that this behavior may reduce innovation by providers of these products to the detriment of society.

Opponents of regulation deny the need for network neutrality regu- lation.4 They argue that regulation is not necessary because network op- erators do not have an incentive to discriminate against independent ap- plications anyway,5 or, alternatively,6 that regulation is harmful because it would reduce network operators’ incentive to upgrade their networks in the future.7

This paper aims at assessing the economic merits of network neu- trality regulation. To this aim, the paper applies insights from game the- ory, industrial organization, antitrust, evolutionary economics and man- agement strategy to analyze network operators’ incentives to discriminate, the impact of potential discriminatory behavior on innova- tion and social welfare, and the costs of regulation. By focusing on the economic merits of network neutrality, the paper complements theoreti- cal approaches that base calls for network neutrality regulation on non-

4. See, e.g., Ex Parte Submission of the National Cable & Telecommunications Asso- ciation to the Declaratory Ruling & Notice of Proposed Rulemaking in Inquiry Concerning High-Speed Access to the Internet, CS Dkt. No. 02-52 (Sept. 8, 2003), http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6514882243. For opponents of nondiscrimination rules in the scientific arena, see, for example, Bruce M. Owen & Gregory L. Rosston, Local Broadband Access: Non Nocere or Primum Processi? A Property Rights Approach (Stanford Law Sch., John M. Olin Program in Law and Econ., Working Paper No. 263, 2003); Christopher S. Yoo, Beyond Network Neutrality, 19 HARV. J.L. & TECH. 1 (2005) [hereinafter Yoo, Beyond Network Neutrality]; Christopher S. Yoo, Network Neutrality and the Economics of Congestion, 94 GEO. L.J. 1847 (2006) [hereinafter Yoo, Economics of Congestion]; Christopher S. Yoo, Would Mandating Broadband Network Neutrality Help or Hurt Competition? A Comment on the End-to-End Debate, 3 J. ON TELECOMM. & HIGH TECH. L. 23 (2004) [hereinafter Yoo, Mandating Network Neutrality]; J. Gregory Sidak, A Consumer-Welfare Approach to Network Neutrality Regulation of the Inter- net, 2 J. OF COMPETITION L. & ECON. 349 (2006), available at http://jcle.oxfordjournals.org/cgi/reprint/2/3/349.pdf. 5. There are two representative examples of this view in the context of the debate over open access to broadband networks. See James B. Speta, Handicapping the Race for the Last Mile?: A Critique of Open Access Rules for Broadband Platforms, 17 YALE J. ON REG. 39 (2000) [hereinafter Speta, Handicapping]; James B. Speta, The Vertical Dimension of Cable Open Access, 71 U. COLO. L. REV. 975 (2000) [hereinafter Speta, Vertical Dimension]. On the open access debate, see infra note 31. 6. Both arguments are mutually exclusive. If network owners do not have an incentive to discriminate against independent applications anyway, the imposition of a network neutrality regime that prevents such discrimination will not reduce their profits. If it does not reduce their profits, however, it cannot reduce their incentives to invest in upgrades of their network infrastructure in the future. 7. For a representative example of this view, see Adam D. Thierer, “Net Neutrality” Digital Discrimination or Regulatory Gamesmanship in Cyberspace? (Cato Inst., Policy Analysis No. 507, 2004).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 333

economic rationales.8 Throughout this paper, the term “network neutrality rules” refers to

non-discrimination rules that forbid operators of broadband networks to discriminate against third-party applications, content or portals (“inde- pendent applications”) and to exclude them from their network. This terminology captures the common rationale behind the various network neutrality proposals before Congress and the FCC – to design rules that prevent network operators and ISPs from using their power over the transmission technology to negatively affect competition in complemen- tary markets for applications, content and portals.9 By contrast, network neutrality opponents sometimes use a much broader definition of net- work neutrality that includes mandating interconnection, non- discrimination, rate regulation and the adoption of standardized protocol interfaces such as TCP/IP.10 While providing a convenient straw man for

8. E.g., Brett M. Frischmann, An Economic Theory of Infrastructure and Commons Management, 89 MINN. L. REV. 917 (2005); Crawford, supra note 1. For a critical evaluation of such approaches, see Yoo, Beyond Network Neutrality, supra note 4, at 53-57. 9. For an overview of the various proposals, see McCullagh, supra note 2, and Public Knowledge White Paper, supra note 2, at 3-7, 26-27. In addition to the non-discrimination rules discussed in the text, network neutrality proposals often include the right of consumers to attach communication equipment of their choice to the network. See Public Knowledge White Paper, supra note 2, at 3-7, 26-27. Network neutrality regulation is not intended to prevent vertical integration between network providers and application providers (i.e., network provid- ers are allowed to offer applications as well). See Wu, Broadband Debate, supra note 1, at 89.

While calls for network neutrality rules share a common rationale, they differ with respect to how these rules should be implemented. For example, in some proposals, the non- discrimination rules take the form of user rights (to access and use the content and applications of their choice), in others the respective rights are vested in the providers of complimentary products (to offer the application and content of their choice). See, e.g., Public Knowledge White Paper, supra note 2, at 3-7, 26-27. Proposals differ with respect to the exceptions to the non-discrimination rule they include, (i.e., with respect to the cases in which a deviation from the principle of network neutrality is justified). Id. at 27. For example, whether and, if yes, what form of price discrimination should be forbidden under network neutrality regulation, is still an open question. Compare Wu, Network Neutrality, supra note 1, at 151-54 (arguing against price discrimination, if it is based on discrimination between applications), and JONATHAN E. NUECHTERLEIN & PHILIP J. WEISER, DIGITAL CROSSROADS 177 (2005); see also Senate Hearing, supra note 3 (testimony of Prof. Lawrence Lessig), available at http://commerce.senate.gov/pdf/lessig-020706.pdf (arguing against “access tiering,” i.e. “any policy by network owners to condition content or service providers’ right to provide content or service to the network upon the payment of some fee [. . . which is] independent of basic Internet access fee,” id. at 2 note 2, but supporting “customer tiering,” i.e. price discrimination, as long as it is not based on discrimination among content or application providers); but see Sidak, supra note 4, at 83-99 (arguing in favor of allowing access tiering). For specific imple- mentation proposals from the scientific literature, refer to Wu, Broadband Debate, supra note 1, appendix A; Weiser, supra note 1, at 74-84; Atkinson & Weiser, supra note 1, at 55. 10. E.g. Yoo, Beyond Network Neutrality, supra note 4, at 3, 8, 27, 32. As most of Yoo’s arguments about the negative impact of network neutrality are based on the negative impact of measures such as the adoption of standardized interfaces that are not part of the net- work neutrality regime discussed in this paper, his analysis does not carry over to the case of “pure” non-discrimination rules discussed here. See also discussion infra notes 192, 198. For a

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

334 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

attack, this definition goes far beyond what network neutrality propo- nents want to achieve: the measures included in the broad definition con- stitute heavy forms of regulation; by contrast, the non-discrimination rules in network neutrality proposals have been explicitly designed to provide a light form of behavioral regulation that narrowly targets the behavior identified as problematic and is far less intrusive than other forms of regulation such as structural separation or open access regula- tion.11

The analysis proceeds in three steps. Part II explores whether net- work providers have an incentive to discriminate against applications.12 This question has not been examined in detail in the existing literature.13 If, however, network providers do not have such an incentive, there is no need for regulation.14

Whether exclusionary conduct in complementary markets is a prof- itable strategy has been hotly debated over the years. Today, most schol- ars agree that a monopolist in a primary market does not generally have an incentive to exclude its competitors from a secondary, complementary

critical appraisal of Yoo’s work on network neutrality, see Frischmann & van Schewick, supra note 1; Herman, supra note 1. 11. See, e.g., Weiser, supra note 1, at 48, 74, 78-80; Wu, Network Neutrality, supra note 1, at 145-49. By contrast, Yoo derives its definition from statements of network neutral- ity proponents with respect to Internet policy in general, not to network neutrality in particular, Yoo, Beyond Network Neutrality, supra note 4, at 3. 12. See infra Part II. 13. For a similar assessment with respect to network neutrality proponents, see Weiser, supra note 1, at 74-75. For an example of the treatment of the question by a network neutrality proponent, see Wu, Broadband Debate, supra note 1, Part II.B. For an example of the treat- ment of the question by a network neutrality opponent, see Yoo, Beyond Network Neutrality, supra note 4, Part II. (arguing that network neutrality proponents’ focus on safeguarding com- petition in the markets for application and content is misplaced without examining whether there is indeed a threat of discrimination) and at 60-61 (arguing that competition in the broad- band market “should remain sufficiently robust to ameliorate concerns of anticompetitive ef- fects” without covering specific motivations for discrimination); Yoo, Economics of Conges- tion, supra note 4 (manuscript at 49-50). While Farrell and discuss exceptions to the “one monopoly rent” argument in detail, their analysis is not specifically targeted at the economic relationships relevant in the network neutrality context. Joseph Farrell & Philip J. Weiser, Modularity, Vertical Integration, and Open Access Policies: Towards a Convergence of Anti- trust and Regulation in the Internet Age, 17 HARV. J.L. & TECH. 85, 114 (2003). Similarly, while participants in the open access debate have explored the incentives of network providers to exclude independent ISPs from their network, their analysis focuses on the competitive rela- tionships between the operators of physical networks on the one hand and unaffiliated ISPs that may also offer applications, content or portals on the other hand. E.g., Speta, Handicap- ping, supra note 5; Speta, Vertical Dimension, supra note 5 (both denying an incentive to ex- clude); Daniel L. Rubinfeld & Hal J. Singer, Vertical Foreclosure in Broadband Access?, 49 J. INDUS. ECON. 299 (identifying an incentive to exclude). By contrast, network neutrality rules focus on the competitive relationships between the operators of physical networks and/or ISPs on the one hand and unaffiliated providers of complementary applications, content and portals on the other hand. Thus, the open access literature is not directly applicable to the network neutrality debate. 14. For a qualification of this assessment, see infra notes 27-28 and accompanying text.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 335

market – the well known “one monopoly rent” argument.15 There are known exceptions to this rule, but these rarely apply. As a result, when analyzing allegations of exclusionary conduct in a complementary mar- ket, most scholars intuitively assume there will not be a problem, in par- ticular if the excluding actor faces competition in the primary market.

The results of the analysis challenge this intuition in several ways: First, Part II identifies exceptions to the “one monopoly rent” argu-

ment that have not been previously thought of, but are quite common in the Internet context.16

Second, the paper shows that some of the known exceptions do in- deed apply in the Internet context.17

Third, researchers commonly assume that discrimination against a complementary product will only be profitable, if the primary good mo- nopolist manages to monopolize the market for the application in ques- tion. The paper shows that this assumption is not necessarily correct. A network operator may have an incentive to discriminate against an appli- cation even if the operator does not manage to drive all independent ap- plications from the corresponding market.18 As a result, researchers commonly underestimate the potential for discriminatory behavior by network providers.

Finally, in line with conventional thinking on the profitability of ex- clusionary conduct, participants in the debate usually share the view that competition in the market for Internet services may be able to mitigate the problem. Two policy proposals, the proposals for facilities-based competition and for open access, are based on this view. The results of Part II contradict this view. The analysis highlights a variety of circum- stances under which a network operator may have the ability and incen- tive to discriminate against independent applications in spite of competi- tion in the market for Internet services.19

Thus, Part II highlights important limitations of the “one monopoly rent” argument in the Internet context that may be relevant beyond the network neutrality debate. In the network neutrality context, it shows that in the absence of network neutrality regulation, there is a real threat of discriminatory behavior that is more severe than is commonly assumed.

Part III analyzes the impact of this threat on innovation in the mar- kets for applications, content and portals (“application-level innova- tion”).20 It shows that the threat of discrimination reduces the amount of

15. See infra Part II.B.1. 16. See infra Part II.B.2. 17. See infra Part II.B.3. 18. See infra Part II.B.4. 19. See infra Part II.C. 20. See infra Part III.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

336 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

application-level innovation by independent producers of complementary products.21 While discrimination increases network providers’ incentives to engage in application-level innovation, this increase cannot offset the reduction in innovation by independent producers.22 Thus, the threat of discrimination reduces the amount of application-level innovation.

Part IV explores the social benefits and costs of network neutrality regulation.23 It shows that the increase in application-level innovation re- sulting from network neutrality rules is socially beneficial.24 On the cost side, network neutrality rules reduce network providers’ incentives to in- novate at the network level and to deploy network infrastructure.25 While regulatory intervention has its own costs, these are not covered in detail. When deciding whether to introduce network neutrality regulation, regu- lators must trade-off the benefits against the costs. The analysis shows that in the context of the Internet, the benefits of network neutrality regu- lation are more important than the costs.26

I. THREAT OF DISCRIMINATION

Calls for network neutrality regulation are based on the assumption that network providers have an incentive to discriminate against unaffili- ated providers of complementary products. If network providers do not have such an incentive, there is no need for regulation. In this case, regu- lation may still serve an educational function and protect customers and providers of independent content, portals and applications from discrimi- natory or exclusionary conduct by “incompetent incumbents”27 that fail to recognize that discrimination is not in their best economic interest.28 Compared to a threat of discrimination due to a real incentive to dis- criminate, this constitutes a considerably weaker basis for regulatory in- tervention.

Network technology gives network providers the ability to discrimi- nate against applications running over their networks or to exclude them from the network. The following part explores, whether network provid- ers have an incentive to actually use this discriminatory power.29 The analysis is based on a stylized model (Section A). As the answer may dif-

21. See infra Part III.A. 22. See infra Part III.B. 23. See infra Part IV. 24. See infra Part IV.A. 25. See infra Part IV.B. 26. See infra Part IV.C. 27. Farrell & Weiser, supra note 13, at 114. 28. Id. at 114-17; Wu, Network Neutrality, supra note 1, at 154-56. 29. There have been various instances of discrimination by network providers in prac- tice, both in the United States and internationally. See generally Public Knowledge White Pa- per, supra note 2, at 16-23.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 337

fer depending on the market structure in the market for Internet services, the analysis proceeds in two steps: In the first step, the network provider is a local monopolist (Section B). In the second step, the network pro- vider competes with at least one other network provider (Section C).

The analysis shows that discrimination is much more likely than is commonly assumed.

A. Stylized Model

Network neutrality rules seek to protect competition in complemen- tary products such as Internet applications, content and portals from anti- competitive behavior by network operators or ISPs. To reflect this goal, the analysis focuses on the competitive interactions between “the net- work” and “applications.”30 Economically, “the network” comprises two distinct layers of economic activity: the operation of physical networks and the provision of Internet access and transport services over these networks. In real life, these activities may or may not be provided by dif- ferent economic actors with differing economic interests. The resulting competitive interactions between network operators and Internet service providers have featured prominently in the debate over “open access” for independent Internet service providers to broadband cable networks in the United States.31 To focus on the specific impact of network neutrality

30. In the context of the four layer model of the Internet Architecture used by the Inter- net Engineering Task Force, “the network” consists of the network layer and the Internet layer, while the application domain consists of the transport layer and the application layer. See, e.g., LARRY L. PETERSON & BRUCE S. DAVIE, COMPUTER NETWORKS: A SYSTEMS APPROACH 27- 30 (3d ed. 2003). 31. The open access debate focuses on the question whether the owners of cable net- works should be required to allow independent Internet service providers to provide Internet access services over their cable networks. Several scholars advocate open access regulation. See Ex parte Submission of Mark A. Lemley & Lawrence Lessig, to the Public Notice, in Ap- plication for Consent to the Transfer of Control of License Licenses from MediaOne Group, Inc. to AT&T Corp., at 1, CS Dkt. No. 99-251 (November 10, 1999), http://gullfoss2.fcc.gov/prod/ecfs/retrieve.cgi?native_or_pdf=pdf&id_document=6010050443 [hereinafter Lemley & Lessig, Ex parte]; Mark A. Lemley & Lawrence Lessig, The End of End-to-End: Preserving the Architecture of the Internet in the Broadband Era, 48 UCLA L. REV. 925 (2001); LESSIG, FUTURE OF IDEAS, supra note 1, at 147-67, 246-49; Francois Bar et al., Access and Innovation Policy for the Third-Generation Internet, 24 TELECOMM. POL’Y 489 (2000); Jim Chen, The Authority to Regulate Broadband Internet Access over Cable, 16 BERKELEY TECH. L.J. 677 (2001); Mark N. Cooper, Open Access to the Broadband Internet: Technical and Economic Discrimination in Closed, Proprietary Networks, 71 U. COLO. L. REV. 1011 (2000); Jerry A. Hausman et al., Cable Modems and DSL: Broadband Internet Ac- cess for Residential Customers, 91 AM. ECON. REV. 302 (2001) [hereinafter Hausman et al., Cable Modems]; Jerry A. Hausman et al., Residential Demand for Broadband Telecommunica- tions and Consumer Access to Unaffiliated Internet Content Providers, 18 YALE J. ON REG. 129 (2001) [hereinafter Hausman et al., Residential Demand]; William P. Rogerson, The Regu- lation of Broadband Telecommunications, the Principle of Regulating Narrowly Defined Input Bottlenecks, and Incentives for Investment and Innovation, 2000 U. CHI. LEGAL F. 119; Daniel L. Rubinfeld & Hal J. Singer, Open Access to Broadband Networks: A Case Study of the

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

338 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

rules, the following analysis abstracts from these issues and treats these players as a single economic entity called the “network provider.”

The analysis will be based on the following stylized model: for a given physical network, Internet access and transport services and the operation of the network infrastructure are provided by the same eco- nomic entity, the “network provider.” The corresponding service will be called “Internet service.” The network is assumed to provide the same general functionality as the Internet in that it enables computers attached to distinct physical, but interconnected networks to communicate. Con- trary to the original Internet,32 but similar to networks today, the network is application-aware and can control the execution of applications run- ning over its network. Today, technology is available that enables net- work operators and ISPs to distinguish between the different applications using the network and to control their execution.33 For example, network

AOL/Time Warner Merger, 16 BERKELEY TECH. L.J. 631 (2001); Rubinfeld & Singer, supra note 13. Several experts also opose open access regulation. See John E. Lopatka & William H. Page, Internet Regulation and Consumer Welfare: Innovation, Speculation, and Cable Bun- dling, 52 HASTINGS L.J. 891 (2001); Glen O. Robinson, On Refusing to Deal With Rivals, 87 CORNELL L. REV. 1177 (2002); Speta, Handicapping, supra note 5; Speta, Vertical Dimension, supra note 5; Glenn A. Woroch, Open Access Rules and the Broadband Race, 2002 L. REV. M.S.U.-D.C.L. 719 (2002); Christopher S. Yoo, Vertical Integration and Media Regulation in the New Economy, 19 YALE J. ON REG. 171 (2002). 32. In the original Internet, the network was application-blind, (i.e., it was unable to distinguish between the applications running over the network). Consequently, network opera- tors were unable to affect the execution of specific applications, shielding independent applica- tion developers from strategic behavior by network operators.

The application-blindness was the result of following the broad version of the end-to-end arguments during the design of the Internet, Barbara van Schewick, Architecture and Innova- tion: The Role of the End-to-End Arguments in the Original Internet 101-03 (Ph.D. disserta- tion, Technical University Berlin 2005, MIT Press forthcoming 2008). This design principle requires that the lower layers of the network be as general as possible, while all application- specific functionality is concentrated at higher layers at end hosts. (There are two versions of the end-to-end arguments: a narrow version, which was first identified, named and described in a seminal paper by Saltzer, Clark and Reed in 1981. Jerome H. Saltzer et al., End-to-End Arguments in System Design, 1981 2ND INT’L CONF. ON DISTRIBUTED COMPUTING SYS. 509 (a revised version of paper was later published as Jerome H. Saltzer et al., End-to-End Argu- ments in System Design, 2 ACM TRANSACTIONS ON COMPUTER SYS.S 277 (1984)). A broad version was the focus of later papers by other authors. See, e.g., David P. Reed et al., Com- mentaries on “Active Networking and End-to-End Arguments”, 12 IEEE NETWORK 69, 69 (1998); Marjory S. Blumenthal & David D. Clark, Rethinking the Design of the Internet: The End-to-End Arguments vs. the Brave New World, 1 ACM TRANSACTIONS ON INTERNET TECH. 70, 71 (2001). While both versions have shaped the original architecture of the Internet, only the broad version is responsible for the application-blindness of the network.) For a detailed analysis of the two versions and their relationship to the architecture of the Internet, see van Schewick, supra, at 87-129. 33. See, e.g., Cisco Systems, Inc., Network-Based Application Recognition and Distrib- uted Network-Based Application Recognition, http://www.cisco.com/en/US/products/ps6350/products_configuration_guide_chapter09186a0 080455985.html (last visited Sept. 30, 2006). This technology violates the broad version of the end-to-end arguments, but as the end-to-end arguments are just a design principle, there is

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 339

providers can slow down selected applications or content, speed them up or exclude them from the network completely.

In the analysis of Section B, the network provider is a local mo- nopolist.34 The size of its footprint relative to the size of the nationwide network may differ. In the extreme case, the network provider owns the nationwide network and has a nationwide monopoly in the provision of Internet services.

In Section C, the network provider competes with at least one other network provider.

The network provider also offers products in the market for applica- tions, content or portals.35 These products may be offered in two differ- ent ways:

In the first case, the complementary product is offered to consumers nationwide. Thus, if the size of the provider’s footprint is smaller than the nationwide network, the product in question is not only offered to customers of its Internet services, but also to consumers living outside its footprint. A product that is offered this way will be called an affiliated product.

Alternatively, the network provider may only offer the product to customers of its Internet service. If the size of the provider’s footprint is smaller than the nationwide network, consumers outside its footprint will not be able to use or buy the product. This kind of product will be re- ferred to as proprietary product.

For a particular product, the two ways of offering the product are mutually exclusive.

This division reflects the way in which network providers’ comple-

nothing that forces technology to comply with it. See van Schewick, supra note 32, at 101-03. 34. See Applications for Consent to the Transfer of Control of Licenses and Section 214 Authorizations by Time Warner Inc. and America Online, Inc., Transferors, to AOL Time Warner Inc., Transferee, Memorandum Opinion & Order, 16 FCC Rcd. 6,547, ¶ 74 (2001) [hereinafter AOL Memorandum Opinion & Order] (“The relevant geographic markets for resi- dential high-speed Internet access services are local. That is, a consumer’s choices are limited to those companies that offer high-speed Internet access services in his or her area, and the only way to obtain different choices is to move. While high-speed ISPs other than cable opera- tors may offer service over different local areas (e.g., DSL or wireless), or may offer service over much wider areas, even nationally (e.g., satellite), a consumer’s choices are dictated by what is offered in his or her locality.”) See also Hausman et al., Residential Demand, supra note 31, at 135 (“From a consumer’s perspective, the relevant geographic market is local be- cause one can purchase broadband Internet access only from a local residence. Stated another way, a hypothetical monopoly supplier of broadband Internet access in a given geographic market could exercise market power without controlling the provision of broadband access in neighboring geographic markets”). 35. Thus, the analysis assumes that the network provider is vertically integrated into the provision of at least some applications. Vertical integration, however, is not the only case to which the analysis applies. A similar analysis applies to other forms of close vertical relation- ships between the network provider and a provider of complementary products such as partial integration, partial equity investments, long-term contracts, or other forms of close affiliation.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

340 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

mentary products are offered in today’s Internet market. For example, AOL offers MapQuest, AOL Moviefone or its instant messenger to any- body using the Internet.36 Similarly, AOL’s portal is available both bun- dled with Internet service and separately.37 By contrast, T-Online, the dominant German Internet provider, offers its portal only bundled with its Internet service.

The subsequent analysis does not further examine the choice of product provisioning, but takes the result as given.

B. Network Provider is Monopolist in the Market for Internet Services

Economic theory predicts that a network operator that has a monop- oly in the market for Internet services does not generally have an incen- tive to discriminate against independent applications (Section 1). There are known exceptions to this rule, but there is considerable debate over whether these apply in the Internet context. The following analysis shows that the threat of discrimination is more severe than is commonly assumed. First, there are more exceptions than have been previously identified (Section 2). Second, some of the known exceptions may be more relevant in the Internet context than is commonly assumed (Section 3). Third, discrimination may be a profitable strategy, even if the net- work provider does not manage to drive independent applications from the market (Section 4).

1. No General Incentive to Discriminate

According to the “one monopoly rent” theory, a monopolist has no incentive to monopolize a complementary product market, if the com- plementary product is used in fixed proportions38 with the monopoly good and is competitively supplied.39

36. Time Warner, Inc., Time Warner Businesses: AOL, Aug. 2, 2006, http://www.timewarner.com/corp/businesses/detail/aol/index.html. 37. Alan Breznick, AOL Shifts Broadband Strategy, CABLE DATACOM NEWS, Jan. 1, 2003, http://www.cabledatacomnews.com/jan03/jan03-3.html. 38. If the two goods are used in variable proportions, the monopolist may have an in- centive to monopolize the complementary market, as this creates greater flexibility in its rela- tive pricing of both components. Through appropriate pricing, the monopolist may be able to extract more surplus from consumers. If it needs a monopoly over both products to price dis- criminate in this fashion, monopolizing the second market will increase its profits. See, e.g., Janusz A. Ordover et al., Nonprice Anticompetitive Behavior by Dominant Firms toward the Producers of Complementary Products, in ANTITRUST AND REGULATION: ESSAYS IN MEMORY OF JOHN J. MCGOWAN 119 (Franklin M. Fisher ed., 1985). 39. See, e.g., ROBERT H. BORK, THE ANTITRUST PARADOX; A POLICY AT WAR WITH ITSELF 372-75 (Free Press 1993) (1978); RICHARD A. POSNER, ANTITRUST LAW 198-99 (2d ed. 2001).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 341

In this case, there is only one final product, and, therefore, only one monopoly profit available in the market for the final product. The mo- nopolist can extract the complete monopoly profit through its pricing of the monopoly good, and does not gain additional profits by monopoliz- ing the complementary good.

This line of reasoning suggests that the monopolist need not mo- nopolize the secondary market to extract the entire monopoly rent and therefore has no incentive to drive rivals from that market.

Moreover, economists note that the monopolist may benefit from the presence of independent producers in the complementary product market, implying that the monopolist will welcome, not exclude inde- pendent producers of complementary products. This argument has been labeled “internalizing complementary efficiencies (ICE).”40

If the presence of independent producers of complementary prod- ucts generates additional surplus, the monopolist may be able to capture some of that surplus through its pricing of the primary good. In this case, the monopolist will earn greater profits when its rivals are in the market than when they are not. In this case, the monopolist does not wish to steal sales in the secondary market, but takes its profits by charging a higher price for the primary good.41

Whether the presence of independent producers generates additional surplus, depends on the structure of consumer preferences and on factors such as the intensity of competition in the complementary market or the degree of differentiation in the complementary market.42

40. Farrell & Weiser, supra note 13, at 89. 41. See, e.g., Michael D. Whinston, Tying, Foreclosure, and Exclusion, 80 AM. ECON. REV. 837, 840, 850-52 (1990); Joseph Farrell & Michael L. Katz, Innovation, Rent Extraction, and Integration in Systems Markets, 48 J. INDUS. ECON. 413 (2000); Farrell & Weiser, supra note 13, at 103. 42. As the intensity of competition increases, prices are driven down to marginal costs. Due to the complementarity between both products, the monopolist benefits from lower prices in the complementary market. The lower prices in the complementary market, the higher de- mand (if demand is responsive to price) or consumer surplus (if demand is inelastic), and, con- sequently, the higher the profits that can be extracted in the primary market. Id.

Given the complementarity between both markets and appropriate consumer preferences, an increase in the quality or variety of complementary goods will increase consumers’ valua- tion of the primary good. For example, consumer surplus rises, if a rival enters with a differen- tiated complementary product and some consumers prefer that product, e.g., Whinston, supra note 41, at 850-52; Dennis W. Carlton & Michael Waldman, The Strategic Use of Tying to Preserve and Create Market Power in Evolving Industries 11 (George J. Stigler Ctr. for the Study of the Econ. and the State, Graduate Sch. of Bus., Univ. of Chicago, Working Paper No. 145, 2000), available at http://gsbwww.uchicago.edu/research/cses/WorkingPapersPDF’s/145.pdf. The value consum- ers derive from greater variety may well differ depending on the type of complementary prod- uct. For example, consumers may value the fifth teleconferencing application less than the fifth multiplayer online game.

In general, two goods are complements if a decrease in the price of one increases the de-

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

342 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

While the “one monopoly rent” theory argues that exclusionary conduct in the complementary market will not increase the monopolist’s profits, the “internalizing complementary efficiencies” theory suggests that such conduct may even reduce its profits.

Recent research shows that this line of reasoning is incomplete: Contrary to the assumptions of the “one monopoly rent” argument, there are cases in which the monopolist profits from monopolizing the com- plementary market. In these cases, the monopolist may profit from the presence of independent producers in the complementary market, but the loss of these profits may be more than offset by the gains associated with discriminating in the complementary market. In other words, although the monopolist may profit from the presence of independent producers in the complementary market, it may profit even more by excluding them from the market. In this case, the monopolist will engage in exclusionary conduct, if the associated profits are larger than the associated costs.43

2. New Exceptions

The following section highlights three exceptions that have not been previously considered. In the first exception, the complementary product is a source of outside revenues that the monopolist cannot extract in the primary market. In the second exception, which is a variant of the first, only the monopolist’s complementary product is a source of outside revenue which is lost when rival producers of the product make the sales. This exception is particularly relevant in the Voice over IP (VoIP) con- text. In the third exception, the exclusionary conduct in the complemen- tary market preserves a legally acquired monopoly in the complementary market.

The following analysis sets out the theories underlying these excep- tions, highlights the conditions under which they apply and shows that these conditions may well be met in the Internet context.

2.1. Complementary Product Source of Outside Revenue

a) Theory

A monopolist in the primary market may be unable to extract the maximum possible profit through its sales of the primary good, if some of the revenue in the complementary market comes from outside

mand for the other. HAL R. VARIAN, INTERMEDIATE MICROECONOMICS; A MODERN APPROACH 112 (5th ed. 1999). 43. See, e.g., Whinston, supra note 41, at 850-52, 855.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 343

sources.44 In conventional markets, firms typically derive their revenue from

sales of products or from fees for the provision of services. Firms also have the option of following the example of the media: they offer value to their customers, but at least partly charge third parties such as adver- tisers. In other words, a part of their revenue stems from selling access to their customers to interested third parties. In the extreme case, consumers get a firm’s product or service for free, while all of the firm’s revenue comes from outside sources.

If firms in the complementary market derive some of their revenue from outside sources, a monopolist in the primary market may be unable to earn the maximum possible profit unless it monopolizes the comple- mentary market as well. To see this, consider the following example: suppose that firms in the complementary market offer their product or service for free and make all their revenue from selling access to their customers to third parties.

Usually, the monopolist can use a variety of tactics to extract or “squeeze” revenue from its rivals: A common set of tactics forces rival producers of the complementary good to lower the quality-adjusted price of their product.45 This increases the consumer surplus available for ex- traction in the primary market. In the example, the price of complemen- tary products already equals zero; thus, these tactics are not feasible.

In another tactic, the monopolist threatens to exclude a rival from the complementary market, unless the rival pays an access charge.46 To be able to apply this tactic, the monopolist must have the power to ex- clude its rivals, for example due to intellectual property rights or because rivals’ access to the primary good requires the monopolist’s cooperation. While this mechanism enables the monopolist to extract its rivals’ out- side revenue, the monopolist may still earn less than if it excludes its ri- vals, monopolizes the complementary product market and captures all outside revenue directly: first, by monopolizing the complementary mar- ket, the monopolist gains a monopoly in the market for access to the us- ers of its primary good. As a result, it will be able to charge higher prices (per customer) for access to its customers than competing producers of complementary products.47 Second, due to its relationship with consum-

44. This theory is new and has not been covered by the existing literature. 45. For an overview of such tactics, see, e.g., Farrell & Katz, supra note 41, at 414-15. 46. See, e.g., id. at 422. 47. Ultimately, this will harm consumers, as firms will pass on at least some of the in- creased costs to their customers. For example, higher advertising fees will ultimately lead to higher prices for the goods that are advertised. See Rubinfeld & Singer, supra note 13, at 316; Jeffrey K. MacKie-Mason, An AOL/Time Warner Merger Will Harm Competition in Internet Online Services 23 (October 17, 2000) (Report submitted to the U.S. Federal Trade Commis- sion), http://www-personal.umich.edu/~jmm/papers/aol-tw00-public.pdf.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

344 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

ers in the primary market, the monopolist may have information about its consumers that enables it to charge higher prices to third parties.48 Third, even if the per customer prices charged to third parties stay the same, the monopolist’s profits will be lower in the presence of rivals due to the costs of negotiating and administering the access fees.

Thus, the monopolist will have an incentive to exclude its rivals from the complementary market, if the gains from directly capturing the outside revenue more than offset the reduction in profits that results from the reduction in complementary goods variety.

b) Application to the Internet

In the market for Internet content, portals and applications, firms of- ten derive at least some of their revenue from outside sources by selling access to their customers to advertisers or online merchants.49

In the hypothetical network that is the focus of this analysis, the monopolist can extract at least some of its rivals’ outside revenue: the network enables the monopolist to exclude applications from the net- work. Thus, the monopolist can condition the “access” of rivals’ products and services on the payment of an access fee that captures some or all of its rivals’ outside revenue. That this is not a mere theoretical possibility shows the practice of cable network owners in the United States. Unaf- filiated Internet service providers who want to offer their service over a cable network have to pay a fixed fee per customer. In addition, the cable network owner receives a portion of the outside revenue that the Internet service provider earns on that customer.50

While the monopolist is able to capture some or all of its rivals’ out- side revenue by threatening exclusion, its outside revenue will be higher if it excludes its rivals and collects the outside revenue directly.

First, selling access to one large group of customers as a whole may yield substantially more revenue than selling access to subgroups of that group separately. This is obvious, if the monopolist network provider manages to monopolize the market in which access to its Internet service customers is sold.51

Second, through its billing relationship with customers of its Inter-

48. See, e.g., CARL SHAPIRO & HAL R. VARIAN, INFORMATION RULES: A STRATEGIC GUIDE TO THE NETWORK ECONOMY 34-35 (1999). 49. ALLAN AFUAH & CHRISTOPHER L. TUCCI, INTERNET BUSINESS MODELS AND STRATEGIES; TEXT AND CASES 56 (2001); SHAPIRO & VARIAN, supra note 48, at 162-63. 50. See Seth Schiesel, New Economy: A New Model for AOL May Influence Cable’s Future, N.Y. TIMES, Aug. 26, 2002, at C1 (discussing a contract between AOL and AT&T Comcast). 51. See, e.g., Rubinfeld & Singer, supra note 13, at 316; MacKie-Mason, supra note 47, at 23. This remains true even if the monopolist does not manage to drive its rivals from the market completely. See the analysis infra Part II.B.4.2.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 345

net service, the network owner has data on customer demographics that enables it to charge higher advertising fees or commissions for online sales than many of its rivals in the market for Internet content, portals and applications.52

Finally, due to the potentially large number of complementary products, negotiating and administering the access charges for unaffili- ated content, applications and portals may be prohibitively expensive. In any event, these transaction costs will further decrease the monopolist’s profits in the presence of rivals.

Thus, if firms in the market for a particular type of application, con- tent or portal derive some of their revenue from outside sources, a mo- nopolist in Internet services may have an incentive to monopolize that market in order to capture all outside revenue available in that market di- rectly.

2.2. Monopolist’s Complementary Product Source of Outside Revenue

a) Theory

In the scenario outlined above, only the network provider, not its ri- vals in the complementary market can realize higher outside revenues. As a result, letting rivals make the sales and extracting the outside reve- nue from them is less profitable than making the sales directly.

The following exception is a variant of this line of reasoning. The network provider’s offering is a source of outside revenue; the rivals’ of- fering does not provide this revenue. Thus, this revenue is lost if rivals make the sales. As a result, the network provider has an incentive to make as many sales as possible directly.

b) Application to the Internet53

Consider a local phone company that offers broadband Internet ser- vices over its network. Independent companies such as Vonage or Skype offer Voice over IP (VoIP) services to customers of this network pro- vider. As the costs of long-distance calls using VoIP are usually consid- erably lower than the costs of long-distance calls using the conventional telephone service, those of the network provider’s customers using VoIP will place less long-distance calls using the network provider’s legacy

52. Even if those rivals require consumers to register before using their product or ser- vice, they have no way to verify the information, unless they require payment; in this case, they can verify the information as part of the billing process. See SHAPIRO & VARIAN, supra note 48, at 34-35; MacKie-Mason, supra note 47, at 11. 53. Thanks to Robert Pepper for highlighting this example.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

346 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

telephone service. To the network provider, conventional long-distance services are a

source of outside revenue that is not similarly available to the providers of VoIP services. In the US, local phone companies are paid so-called access charges by long-distance providers for every long-distance call they originate or terminate. As access charges were traditionally intended to implicitly cross-subsidize local telephone service, regulators have mostly set these access charges significantly above the costs of originat- ing or terminating long-distance calls. Thus, for many local phone com- panies, access charges are an important source of revenue.54

Independent VoIP providers threaten the source of this revenue: The more of the network provider’s telephone customers place their long- distance calls using VoIP, the less access charges the network provider will receive. If independent VoIP providers are excluded from the net- work and the network provider does not offer VoIP itself,55 customers are forced to make their long-distance calls using the conventional tele- phone service. Thus, exclusion in the VoIP market serves to preserve the network provider’s current profits.56

It is not surprising that the first publicly documented incident of VoIP blocking involved a rural telephone company.57 For rural phone companies, access charges constitute a substantial portion of their reve- nue. Thus, they have a particularly high incentive to protect this revenue.

54. See NUECHTERLEIN & WEISER, supra note 9, at 195, 204, 294. 55. The access charge is lost if the call is placed using VoIP, regardless of whether VoIP is provided by the network provider or by an independent provider. Thus, the network provider has an incentive not to have VoIP used on its network at all. 56. In the example discussed in the text, the existence of the outside revenue is the re- sult of regulation that requires long-distance providers to pay above-cost access charges to lo- cal phone companies. Whether local phone companies that are local monopolists in the market for Internet services (this assumption holds throughout Section II.B) would also have an incen- tive to block VoIP in the absence of such regulation, is more difficult to determine. 57. In February 2005, Vonage, a US VoIP provider, complained to the Federal Com- munications Commission that its Internet telephony application was being blocked by Madison River Communications, a rural, local telephone company based in North Carolina. After a short investigation, Madison River and the FCC entered into a consent decree in March 2005. Madison River agreed to voluntarily pay $15,000 as well as to stop blocking VoIP applica- tions; the FCC terminated the investigation. See Madison River Communications, LLC and Affiliated Companies, Order, 20 FCC Rcd. 4,295 (2005); Ben Charny, Vonage Says Broad- band Provider Blocks Its Calls, CNET News.com, http://news.com.com/2100-7352_3- 5576234.html (last modified Feb 14, 2005); Declan McCullagh, Telco Agrees to Stop Blocking VoIP Calls, CNET News.com, http://news.com.com/2100-7352_3-5598633.html (last modi- fied Mar 3, 2005); Madison River Communications, Who We Are, at http://www.madisonriver.net/about_us/who_we_are.php (last visited Nov 21, 2006).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 347

2.3. Monopoly Preservation in the Complementary Market

a) Theory

The monopolist may also use its monopoly over the primary good to protect a monopoly in the complementary market against dynamic com- petition. In this case, the exclusionary conduct in the complementary market preserves the monopoly in that market.58

For this theory to apply, the following conditions must be met:59 First, the monopolized product is not essential for all uses of the

complementary good (i.e., there are uses of the complementary good that do not require the primary good). Second, the monopolist can prevent its rivals from selling their version of the complementary good to users of the primary good. Third, the complementary market is subject to econo- mies of scale or network effects. Fourth, the monopolist also has a mo- nopoly in the complementary market.

While the first condition explains why the monopolist will want to maintain its monopoly in the complementary market in spite of its mo- nopoly in the primary market, the second and third condition provide the mechanism that enables the monopolist to protect its monopoly in the complementary market.

The first condition provides the motivation for preserving a monop- oly in the complementary market in spite of the monopoly in the primary market: The existence of uses of the complementary good that do not re- quire the primary good deprives the monopolist of its ability to extract all profits through sales of the primary good.

To see this, consider the following example: suppose there is some use of the complementary good that does not require the primary good. As a result, the complementary market consists of two parts: a “systems market” for uses in which the primary good is essential, and a “stand- alone market” for uses that do not require the primary good; consumers in the systems market desire the primary and the complementary good,

58. This theory has not been used as an exception to the “one monopoly rent” argument before. It generalizes from an argument that was used by the Federal Communications Com- mission in the AOL/ Time Warner merger proceeding with regard to instant messaging. AOL Memorandum Opinion & Order, supra note 34, at 6603-29, ¶¶ 128-200; Gerald Faulhaber, Network Effects and Merger Analysis: Instant Messaging and the AOL-Time Warner Case, 26 TELECOMM. POL’Y 311 (2002). See infra Part II.B.2.3.b). 59. The structure of the model and the underlying reasoning are parallel to the “primary good not essential” case outlined infra Part II.B.3.1. Whinston, supra note 41, at 854-55. However, in the “primary good not essential” case, the monopolist takes advantage of econo- mies of scale and network effects in the complementary market to extend its monopoly to the complementary market by excluding its rivals from the systems part of the market. In the case under consideration here, the monopolist uses the same mechanism to protect a legally ac- quired monopoly in the complementary market against emerging competition.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

348 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

whereas consumers in the stand-alone market desire only the comple- mentary good.

Suppose there are rival producers of the complementary good. The monopolist can extract all monopoly profits in the systems market through its pricing of the primary good. As consumers in the stand-alone market do not buy the primary good, however, the monopolist does not derive any profit from its rivals’ sales in that market. Moreover, the pres- ence of rivals constrains its ability to price its version of the complemen- tary good in the stand-alone market.

Thus, the monopolist cannot earn monopoly profits in the stand- alone market, unless it has a monopoly in that market. Consequently, keeping competitors out of the complementary market is a prerequisite for preserving current profits.

The second and third condition provide the mechanism that enables the monopolist to preserve the monopoly in the complementary market: In the presence of economies of scale or network effects, the monopolist may be able to drive potential rivals from the complementary market by excluding them from the systems part of the market.

When the second condition is met, the monopolist can deprive rival producers of complementary products of any sales in the systems part of the market.

This behavior does not exclude rivals from the stand-alone market. Given economies of scale60 in the complementary market, the remaining sales to customers in the stand-alone market may not suffice to reach an economically efficient scale. Thus, being excluded from the systems part of the market, rivals may be forced to exit the stand-alone market as well.

Similarly, in the presence of network effects61 in the complementary

60. Economies of scale exist, if an increase in output causes long run average total costs to decrease. In other words, the more output is produced, the lower the cost per unit. E.g., ROBERT E. HALL & MARC LIEBERMAN, ECONOMICS; PRINCIPLES AND APPLICATIONS 177-78 (2d ed. 2001). For example, economies of scale exist, if fixed costs are large relative to mar- ginal costs. In this case, an increase in output allows the firm to spread the fixed costs of pro- duction over greater amounts of output, lowering the costs of unit per output. 61. Network effects exist if the utility an individual customer derives from the con- sumption of a good depends upon, and increases with, the number of other customers who consume products that are compatible with that good. See, e.g., the definition by Michael L. Katz & Carl Shapiro, Network Externalities, Competition, and Compatibility, 75 AM. ECON. REV. 424, 424 (1985) [hereinafter Katz & Shapiro, Network Externalities]. Network effects are covered by a large body of literature. See, e.g., Jeffrey Rohlfs, A Theory of Interdependent Demand for a Communications Service, 5 BELL J. ECON. & MGMT. SCI. 16 (1974); Paul A. David, Clio and the Economics of QWERTY, 75 AM. ECON. REV. 332 (1985); Joseph Farrell & Garth Saloner, Standardization, Compatibility, and Innovation, 16 RAND J. ECON. 70 (1985); Katz & Shapiro, Network Externalities, supra; Michael L. Katz & Carl Shapiro, Technology Adoption in the Presence of Network Externalities, 94 J. POL. ECON. 822 (1986); Carmen Mat- utes & Pierre Regibeau, “Mix and Match”: Product Compatibility without Network External-

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 349

market, exclusion from the systems part of the market may suffice to drive competitors from the market or into a niche existence. In markets with network effects, the incumbent’s large installed base makes it diffi- cult for new entrants to dislodge the incumbent. Exclusion from the cus- tomers in the systems part of the market makes it even more difficult for new entrants to reach the critical mass of customers necessary to start the positive feedback required to succeed with their product.

Thus, the exclusion of rivals from the systems part of the market enables the monopolist to protect a legally acquired monopoly in the complementary market against emerging competition.

Such a scenario may be particularly relevant, if the complementary market belongs to an R&D intensive industry subject to dynamic or “Schumpeterian” competition.62 Due to the presence of intellectual prop- erty rights, economies of scale or network effects, R&D intensive indus- tries are prone to short run exercise of market power. In other words, competition in these markets often results in a single firm dominating the market. Thus, firms in these industries typically compete “for the mar- ket,” not “within the market.” While firms with market power (the win- ners of the competition) are an inherent feature of such industries, their dominance may be temporary, as rapid technological change and drastic

ities, 19 RAND J. ECON. 221 (1988); Brian W. Arthur, Competing Technologies, Increasing Returns, and Lock-In by Historical Events, 99 ECON. J. 116 (1989); Jeffrey Church & Neil Gandal, Network Effects, Software Provision, and Standardization, 40 J. INDUS. ECON. 85 (1992); Nicholas Economides & Steven C. Salop, Competition and Integration among Com- plements, and Network Market Structure, 40 J. INDUS. ECON. 105 (1992); Joseph Farrell & Garth Saloner, Converters, Compatibility, and the Control of Interfaces, 40 J. INDUS. ECON. 9 (1992); Michael L. Katz & Carl Shapiro, Product Introduction with Network Externalities, 40 J. INDUS. ECON. 55 (1992); Stanley M. Besen & Joseph Farrell, Choosing How to Compete: Strategies and Tactics in Standardization, 8 J. ECON. PERSP. 117 (1994); Michael L. Katz & Carl Shapiro, Systems Competition and Network Effects, 8 J. ECON. PERSP. 93 (1994) [herein- after Katz & Shapiro, Systems Competition]; Nicholas Economides, The Economics of Net- works, 14 INT’L J. INDUS. ORG. 673 (1996); see also SHAPIRO & VARIAN, supra note 48, chap- ters 7-9 (analyzing network effects in the context of information goods); Joseph Farrell & Paul Klemperer, Coordination and Lock-In: Competition with Switching Costs and Network Effects, in 3 HANDBOOK OF INDUSTRIAL ORGANIZATION (forthcoming) (providing a recent survey), available at http://ssrn.com/abstract=917785; Mark A. Lemley & David McGowan, Legal Im- plications of Network Economic Effects, 86 CAL. L. REV. 479 (1998) (analyzing the legal im- plications of network economic effects). For some critical voices, see STAN J. LIEBOWITZ & STEPHEN E. MARGOLIS, WINNERS, LOSERS & MICROSOFT. COMPETITION AND ANTITRUST IN HIGH TECHNOLOGY (rev. ed. 2001); William J. Kolasky, Network Effects: A Contrarian View, 7 GEO. MASON L. REV. 577 (1999). 62. On dynamic or “Schumpeterian” competition, see JOSEPH A. SCHUMPETER, CAPITALISM, SOCIALISM AND DEMOCRACY, 81-86 (Harper Perennial 1975); see also Dennis W. Carlton & Robert H. Gertner, Intellectual Property, Antitrust and Strategic Behavior 19-22 (Nat’l Bureau of Econ. Research, Working Paper No. 8976, 2002); David S. Evans & Richard L. Schmalensee, Some Economic Aspects of Antitrust Analysis in Dynamically Competitive Industries (Nat’l Bureau of Econ. Research, Working Paper No. 8268, 2001); Howard A. She- lanski & Gregory J. Sidak, Antitrust Divestiture in Network Industries, 68 U. CHI. L. REV. 1, 10-15 (2001).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

350 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

innovations may cause demand for their product to collapse: for exam- ple, rivals may come up with a vastly superior product or develop a new product that makes the incumbent’s product obsolete. Thus, incumbents in these industries are primarily constrained by dynamic competition – by the innovation of other firms seeking to replace the existing firm with market power. To avoid being dislodged by rivals, incumbents are forced to innovate themselves.

In the scenario described above, a monopolist could use its market power in the primary market to preserve the legally obtained market power in the complementary market, distorting the dynamic competition for future market power. Instead of innovating to prevent being dis- lodged by competitors, the monopolist could simply exclude its rivals from the systems part of the complementary market, preventing them from reaching the scale or network size necessary to displace the incum- bent.

b) Application to the Internet

The conditions underlying this model may well be present in the Internet context.

First, a specific provider’s Internet service may be non-essential for using applications or accessing content. Consider the market for residen- tial broadband Internet access in the United States.63 Depending on local conditions, the owner of a cable network that provides broadband Inter- net access through its affiliated broadband Internet access provider may well be a local monopolist.64 While this monopolist offers broadband Internet access only in the area covered by its network, it may offer con- tent or applications to Internet users nationwide. In this case, the area covered by its network constitutes the “systems market,” while custom- ers outside its footprint make up the “stand-alone market.”

Such a situation is not uncommon. For example, where it has been able to strike a deal with cable network owners, AOL offers its portal bundled with broadband Internet access. In addition, consumers nation- wide can buy the portal without access, known as the “bring your own access” option.65 Other AOL services such as MapQuest or AOL Movie- fone are also offered to all consumers on the Internet.66 Similarly, if a narrowband access provider has a monopoly with respect to narrowband access, but offers its portal both to its narrowband access customers and

63. The market for broadband Internet access is considered a distinct market from the narrowband access market, see, e.g., AOL Memorandum Opinion & Order, supra note 34, at 6574-77, ¶¶ 68-73; Hausman et al., Residential Demand, supra note 31, at 135-57. 64. See supra note 34. 65. Breznick, supra note 37. 66. See Time Warner, Inc., supra note 36.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 351

to anybody on the Internet, the narrowband access service will be non- essential for customers accessing the portal via broadband access ser- vices.67

Second, in the hypothetical network that forms the basis of the analysis, the monopolist can technically exclude rivals’ applications, content or portals from running over its network. As a result, the mo- nopolist’s Internet service customers (the consumers in the systems mar- ket) are unable to access or use these products. Thus, rivals are deprived of any sales in the systems part of the market.

Third, the markets for software applications, Internet content and portals are all subject to significant economies of scale. The development of these products and services is characterized by large fixed costs, while the marginal costs of production and distribution over the Internet are very small. Thus, the marginal cost of production68 is very low relative to the average cost of production,69 resulting in significant economies of scale.70

In addition, many software applications are subject to direct or indi- rect network effects.71 For example, a communication service like instant

67. Scott Beardsley et al., Making Sense of Broadband, MCKINSEY Q., Issue 2, 2003 at 78-87 (showing that “so far, [. . .] faster and better access to the Internet is the sole killer appli- cation of broadband”) Thus, the scenario described in the text may be quite common. See also Farrell & Weiser, supra note 13, at 119. 68. The marginal cost of production is the incremental cost of producing an additional unit of the good. Thus, the marginal cost of production does not include the costs of product development, e.g., HALL & LIEBERMAN, supra note 60, at 168-69. In the case of software ap- plications, Internet content and portals, the marginal cost of production is the cost of making an additional digital copy of the product, which is typically very low. 69. The average cost of production indicates a firm’s total cost per unit of output. In other words, it denotes the total cost associated with a particular product divided by the quan- tity of output produced. Thus, contrary to the marginal cost of production which does not in- clude the cost of developing the first unit of the product, the average cost of production in- cludes the cost of development divided by the total number of copies. E.g., id. at 168. 70. This cost structure (low marginal costs relative to average costs), which results in significant economies of scale, is generally viewed as a key economic characteristic of the markets for these products. See, e.g., SHAPIRO & VARIAN, supra note 48, at 3-4 (discussing information goods in general); Michael L. Katz & Carl Shapiro, Antitrust in Software Markets, in COMPETITION, INNOVATION, AND THE MICROSOFT MONOPOLY: ANTITRUST IN THE DIGITAL MARKETPLACE 29 (Jeffrey A. Eisenach & Thomas M. Lenard eds., 1999) (discussing software markets), manuscript available at http://faculty.haas.berkeley.edu/shapiro/software.pdf; POSNER, supra note 39, at 245-46 (dis- cussing Internet content, portals and software); MacKie-Mason, supra note 47, at 14 (discuss- ing broadband portals); Rubinfeld & Singer, supra note 13, at 307 (discussing broadband con- tent). 71. Network effects are called “direct network effects,” if the consumption benefits di- rectly result from the size of the network. E.g., Katz & Shapiro, Network Externalities, supra note 61, at 424. “Indirect network effects” exist, if consumer demand for the primary good increases with the variety of complementary goods and services. In this case, network effects arise from supply-side economies of scale in the complementary market: a larger installed base for the primary product allows application developers to spread sunk development costs over a

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

352 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

messenger or Internet telephony is more valuable the more people can be contacted using the service.72 Viewers for multimedia content are subject to indirect network effects:73 The larger the catalogue of content avail- able in a particular format, the more users value owning viewers com- patible with that format. At the same time, content providers are more likely to incur the costs of coding their content in a particular format, the larger the installed base of viewers compatible with that format.

Finally, at least some of these markets are subject to rapid techno- logical change. Not surprisingly, markets for software applications are the canonical example of R&D intensive industries subject to dynamic competition.74

Now consider a network provider that is a local monopolist in Inter- net services and has acquired a dominant position in the nationwide mar- ket for a particular application. Such a provider has an incentive to ex- clude rivals from that market to protect itself from dynamic competition and preserve its monopoly in that market. Whether the monopolist will manage to prevent new entrants from entering the complementary market by excluding them from access to its Internet service customers, depends on the exact size of economies of scale with respect to the product in question, on the strength of any potential network effects and on the size of both the monopolist’s network and the remaining network.

This theory played an important role in the FCC’s evaluation of the merger between AOL and Time Warner. Time Warner owned a number of broadband cable networks; AOL held a dominant position in the mar- ket for instant messaging services and offered its instant messaging pro- gram to consumers nationwide. The FCC was concerned that the merged firm could use its control over broadband cable networks to disadvantage competitors seeking to overturn AOL’s legally acquired monopoly in in- stant messaging services. To alleviate this problem, the FCC approved the merger subject to a condition (among others) that required AOL Time Warner to interoperate with instant messenger competitors prior to offering “advanced” instant messaging services.75

larger potential sales base. Thus, in the presence of economies of scale and free entry into the complementary product market, a larger customer base leads to lower costs and greater variety of complementary products. See, e.g., id. at 424; Katz & Shapiro, Systems Competition, supra note 61, at 99. The existence of direct or indirect network effects is a fundamental economic characteristic of many software markets. See, e.g., EVANS & SCHMALENSEE, supra note 62, at 9-11; Katz & Shapiro, supra note 70. 72. E.g., Faulhaber, supra note 58. 73. E.g., MacKie-Mason, supra note 47, at 16. 74. E.g., EVANS & SCHMALENSEE, supra note 62, at 4-15. 75. AOL Memorandum Opinion & Order, supra note 34, at 6603-29, ¶¶ 128-200. For an in-depth analysis of the economic rationale underlying this condition, see Faulhaber, supra note 58.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 353

3. Relevance of Known Exceptions

There are a number of known exceptions to the “one monopoly rent” argument and to the “internalizing complementary efficiencies” ar- gument outlined above. The following section describes two exceptions that may be relevant in the network neutrality context, but whose rele- vance in the network neutrality context has not been discussed in detail yet.76

In the first exception, the primary good is not essential for all uses of the complementary good, making it impossible for the monopolist to extract all monopoly profits through its pricing of the primary good.

In the second exception, the monopolist excludes competitors from the complementary market in order to protect its monopoly in the pri- mary market.

3.1. Primary Good not Essential

a) Theory

The structure of models in this category,77 and the underlying rea- soning, is similar to the “monopoly preservation in the complementary market” case described above:

First, the monopolist has a monopoly in the primary market and the primary good is not essential (i.e., there are uses of the complementary good that do not require the primary good). Thus, the complementary market consists of a systems market and a stand-alone market. As a re- sult, the monopolist cannot extract all profits through its pricing of the primary good and profits from extending its monopoly to the comple- mentary market.

Second, there is a mechanism that enables the monopolist to ex- clude rival producers of the complementary good from the systems part of the market. Third, the complementary market is subject to economies of scale or network effects.

76. For a more complete overview of known exceptions to the “one monopoly rent” argument, see Farrell & Weiser, supra note 13, at 105-19; van Schewick, supra note 32, at 245-67. 77. The following theory was developed by Whinston, supra note 41, at 854-55, and is widely accepted as an exception to the “one monopoly rent” argument. See, e.g., Dennis W. Carlton, A General Analysis of Exclusionary Conduct and Refusal to Deal: Why Aspen and Kodak Are Misguided, 68 ANTITRUST L.J. 659, 667-68 (2001); Dennis W. Carlton & Michael Waldman, The Strategic Use of Tying to Preserve and Create Market Power in Evolving In- dustries, 33 RAND J. ECON. 194, 195 (2002); Jay Pil Choi & Chris Stefanadis, Tying, Invest- ment, and the Dynamic Leverage Theory, 32 RAND J. ECON. 52, 55 (2001); Whinston, supra note 41, at 71. For a detailed application of this theory in the context of the open access debate, see Rubinfeld & Singer, supra note 13. See also Farrell & Weiser, supra note 13, at 119.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

354 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

Given economies of scale in the complementary market, the mo- nopolist can force its rivals to exit the stand-alone market by excluding them from the systems part of the market, extending its monopoly to the complementary market.78

Similarly, in the presence of network effects79 in the complementary market, exclusion from the systems part of the market may suffice to drive competitors from the market or into a niche existence:

If the benefits derived from a larger network are large relative to the benefits of product differentiation in the network good, competition be- tween two incompatible technologies will usually result in a single tech- nology dominating the market.80 The reason is that network effects give rise to strong positive feedback in technology adoption: other things be- ing equal, consumers derive larger benefits from a larger network. As the larger network is more attractive, more consumers will join that network, making it even more valuable, leading to even more consumers joining the network. Once this positive feedback loop sets in, the affected tech- nology will quickly pull away from its rivals in market share, ultimately dominating the market. This phenomenon is referred to as “tipping.”81

As small initial advantages may quickly get magnified, small differ- ences, in either perception82 or reality, may determine the outcome of the competition. Therefore, establishing an early lead in installed base83 that is large enough to start the positive feedback loop is an important strat- egy in network markets.84

78. In the “monopoly preservation in the complementary market” case described supra Part II.B.2.3, the monopolist uses this mechanism to protect a legally acquired monopoly in the complementary market against emerging competition. 79. On network effects in general, see supra note 61. On direct and indirect network effects, see supra note 71. 80. Often, competitors will not be driven completely from the market. In particular, some customers with high switching costs or a unique preference for a competitor’s product will prefer to stay with that competitor in spite of the strong network effects associated with the winning technology. See, e.g., Faulhaber, supra note 58, at 329 n.37. 81. “‘Tipping’ occurs when a single provider reaches a critical mass of customers that are so attractive to others that competitors must inevitably shrink, in the absence of interopera- tion.” Id. at 316. 82. In network markets, consumer expectations about the future size of the network play a crucial role in determining the outcome of the competition. This is due to the costs of belong- ing to the losing network: A consumer who has chosen the losing network can either switch to the winner, which may be costly, or suffer from the lower value of a small network. To avoid this situation, the consumer will choose the network that it expects to be the winner. See, e.g., Besen & Farrell, supra note 61, at 118. 83. The installed base is the total number of consumers who have already bought the network good. 84. A substantial lead in installed base is not the only factor that influences the outcome of the competition. Due to the huge benefits of belonging to the winning network, users have a strong desire to choose the technology that will ultimately prevail. Therefore, consumers ex- pectations of who the winner will be are at least as important. Other factors that may influence customers’ expectations and that may therefore result in a competitive advantage are an estab-

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 355

Thus, if the monopolist excludes its rivals from the complementary market, it can capture all customers in the systems market. If the systems market is large enough, the monopolist’s advantage in that market may enable it to reach a critical mass of customers that are so attractive to others that positive feedback sets in, making it impossible for a rival to catch up.

If the presence of rivals increases consumer surplus, the exclusion of rivals may reduce the monopolist’s profits in the systems market.85 In such a case, monopolizing the complementary market increases the mo- nopolist’s profits, if the gain from monopolizing the stand-alone market is larger than the loss resulting from the exclusion of the rival in the sys- tems market.86

If the complementary market is subject to network effects, two ef- fects make it even more likely that exclusion is a profitable strategy:

First, the potential profits from winning the competition between in- compatible technologies are huge, increasing the benefits of exclusion. Imagine a competition between incompatible technologies that are sub- ject to indirect network effects. If the winning standard is protected by intellectual property, the winner can make money on any primary and complementary product that uses the standard. Given the potentially large number of complementary products in markets with indirect net- work effects, licensing fees can lead to substantial profits.87 For example, the winner in the standard competition between competing media player technologies who wins with a proprietary standard protected by intellec- tual property will not only dominate the market for media players, but will also be able to charge licensing fees for every piece of music or video that is encoded for use with the player.

Second, if the complementary product is subject to network effects, the presence of an independent rival in the complementary market does not necessarily increase the monopolist’s profits in the systems market, a fact that reduces the costs of exclusion. If the monopolist’s and the ri- val’s complementary product are incompatible, sales to the rival decrease the size of the network of users of the monopolist’s complementary

lished reputation, a well-known brand name, or ready visible access to capital. Thus, an un- known firm with an early lead may be overtaken by a market leader that enters second, but has a well-known brand name and good reputation. See, e.g., Katz & Shapiro, Systems Competi- tion, supra note 61, at 107. 85. For example, if the rival produces a differentiated product, the rival’s presence cre- ates additional surplus, some of which the monopolist can extract through its sales of the pri- mary good. Thus, the monopolist’s profits in the systems market are increased if its rival is in the market. 86. See Whinston, supra note 41, at 850-52, 855. 87. Due to the cost structure of information products, profits are not even dependent on charging a monopoly price. See the analysis infra Part II.B.4.1.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

356 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

product. As a result, the value users can derive from the monopolist’s complementary product (and the profit the monopolist can extract from them) is lower than the corresponding value if the rival does not make any sales.88

b) Application to the Internet

As has been set out above,89 the conditions underlying this theory are quite common in the Internet context:

Network providers may be local monopolists in the market for Internet services, but offer applications, content or portals to consumers nationwide. Network technology enables network providers to exclude providers of complementary products from access to its Internet service customers. At the same time, the markets for applications, content or por- tals are usually subject to significant economies of scale and, potentially, network effects.

As a result, an Internet service provider may be able to force its ri- vals from the nationwide market (the stand-alone market) by excluding rival portal, content or application providers from access to its Internet service customers (the systems part of the market). Whether exclusion from its Internet service customers suffices to drive its competitors from the nation-wide market90 depends on the exact size of economies of scale with respect to the product in question, on the strength of any potential network effects and on the size of both the monopolist’s network and the remaining network.91

Such a provider will have an incentive to monopolize the market for a particular type of application, content or portal, if the increased profit from additional application, content or portal sales nationwide more than offsets the reduction in broadband access revenues due to the reduction in variety resulting from the exclusion of its rivals with respect to its Internet service customers.92

88. Carlton & Waldman, supra note 77, at 206-07. 89. See supra Part II.B.2.3.b. 90. See Rubinfeld & Singer, supra note 13, at 310-13 (providing a numerical example). Their paper assesses the likelihood of content discrimination (i.e., blocking or degrading the quality of outside content) by a broadband network provider that is vertically integrated into the market for broadband content and portals in the context of the merger between AOL and Time Warner. 91. Even if the monopolist’s footprint is not large enough to force its rivals to exit the market completely, exclusion from a part of the market may put them at a severe competitive disadvantage by forcing them to operate at a less efficient scale or with a smaller network. See the analysis infra Part II.B.4.1. 92. See, e.g., Rubinfeld & Singer, supra note 13, at 310-13.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 357

3.2. Monopoly Preservation in the Primary Market

a) Theory

In the following class of models, exclusionary behavior in the com- plementary market maintains the monopoly in the primary market.93

In models belonging to this category, the monopolist faces potential competition in the primary market. The monopolist can deter entry to the primary market by engaging in exclusionary conduct in the complemen- tary market. Thus, by deterring entry to the primary market, the exclu- sionary behavior in the complementary market preserves the monopoly in the primary market.

Economists have come up with a number of explanations of why exclusionary conduct in the complementary market may be able to deter entry to the primary market. The following analysis will focus on an ex- planation that is particularly relevant in the Internet context: the exclu- sionary behavior in the complementary market harms future competitors in the primary market by depriving them of a source of complementary products.94 As a result, in order to make any sales in the primary market, an entrant to the primary market needs to enter the complementary mar- ket as well (or otherwise secure a sufficient supply of complementary products). If this is significantly more difficult or costly than entering the primary market alone, potential entrants to the primary market may re- frain from entering.

For such a strategy to succeed, two conditions must be met: First, the exclusionary behavior in the complementary market must

deprive a potential entrant to the primary market of a source of comple- mentary products. As a result, the entrant cannot enter the primary mar- ket alone, but must enter both markets at once.

Second, simultaneously entering both markets must be more diffi- cult or costly than the sum of the costs of entering both markets on their own.95 Otherwise, the exclusionary behavior in the complementary mar-

93. On this type of monopoly maintenance in general, see, e.g., Carlton, supra note 77, at 668-71; Farrell & Weiser, supra note 13, at 109-12; Steven C. Salop & R. Craig Romaine, Preserving Monopoly: Economic Analysis, Legal Standards, and Microsoft, 7 GEO. MASON L. REV. 617, 623-24 (1999). For specific models, see, for example, Carlton & Waldman, supra note 77; Choi & Stefanadis, supra note 77. 94. See, e.g., Carlton, supra note 77, at 669-70. 95. E.g., U.S. Dep’t of Justice & FTC, Non-Horizontal Merger Guidelines, § 4.212 (promulgated in 1984 and reaffirmed in 1992 and 1997) (“The relevant question is whether the need for simultaneous entry to the secondary market gives rise to a substantial incremental difficulty as compared to entry into the primary market alone. If the entry at the secondary level is easy in absolute terms, the requirement of simultaneous entry to that market is unlikely adversely to affect entry to the primary market.”), available at http://www.usdoj.gov/atr/public/guidelines/2614.htm [hereinafter Non-Horizontal Merger

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

358 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

ket is unlikely to adversely affect entry to the primary market. Economists have identified four alternative reasons why simultane-

ous entry to both markets may be significantly more difficult or costly than the sum of the costs of entering each market on its own:

-increased cost of capital, -differing economies of scale in both markets, -the uncertainty of innovation, or -the existence of indirect network effects.

Increased Cost of Capital

An entrant that is forced to enter both markets may face an in- creased cost of capital, if it only has experience relevant for operating in one of the markets. If the skills and knowledge necessary to succeed in both markets differ considerably, the increased probability of failure due to his inexperience in one of them may lead lenders to charge a higher rate for the necessary capital. The risk premium will be even larger, if the entrant has to incur huge sunk costs to enter the market. The higher sunk costs, the more costs cannot be recovered in the event of failure.96

Differing Economies of Scale in Both Markets

Entering two markets is more difficult than entering one, if the minimum efficient scale in both markets differs considerably. In this case, the entrant must choose between operating at an inefficiently small size in one market or at a larger than necessary scale in the other. Both strategies may significantly increase the operating costs of the entering firm.97

Uncertainty of Innovation

Given the uncertainty associated with the innovative process, the need to innovate successfully in two markets may decrease the probabil- ity of successful entry. To see this, assume that the probability of inno- vating successfully in one component is k. In this case, the chances of successful innovation in n components are kn. Unless k is close to 1, this is considerably lower than k.98 Thus, the probability of successful inno- vation in n components required to enter into n markets simultaneously is lower than the probability of successful innovation and successful entry

Guidelines]. 96. For an argument along these lines, see Oliver E. Williamson, Assessing Vertical Market Restrictions: Antitrust Ramifications of the Transaction Cost Approach, 127 U. PA. L. REV. 953, 953-93 (1979); Non-Horizontal Merger Guidelines, supra note 95, § 4.212. 97. Non-Horizontal Merger Guidelines, supra note 95, § 4.212. 98. Carlton and Gertner, supra note 62, pp. 23-27; Choi & Stefanadis, supra note 77.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 359

in one component market.

Existence of Indirect Network Effects

If the primary good is subject to indirect network effects99 and any available complementary goods are offered exclusively with the mo- nopolist’s platform, an entrant into the primary market faces a “chicken and egg” problem: due to consumers’ desire for variety in complemen- tary products, consumers prefer a primary good that already offers a large number of complementary goods and services. At the same time, due to economies of scale and sunk costs in complementary product de- velopment, developers of complementary products prefer to develop products for primary goods that already have a large number of users. Thus, “[an entrant into the primary market] either has to offer consumers much lower value or has to incur large sunk costs to develop (or subsi- dize) a wide range of [complementary goods and services] before there is a large user base to purchase them.” 100

b) Application to the Internet

The conditions underlying this theory may well be present in the Internet context.

First, the exclusionary behavior in the complementary market must deprive a potential entrant to the market for Internet services of a source of complementary products.

By excluding rival producers of Internet portals, content and appli- cation from its network, the monopolist network provider may be able to drive its rivals from the nationwide market.

To deprive a potential entrant of a source of complementary prod- ucts, the monopolist needs not only drive rival content and application producers from the market. He also needs to deny access to its own con- tent and applications to consumers outside its network.101 Otherwise,

99. For a definition of indirect network effects, see supra note 71, 351. 100. Richard J. Gilbert & Michael L. Katz, An Economist’s Guide to US v. Microsoft, 15 J. ECON. PERSP. 25, 30 (2001) (referring to operating systems and application programs). Un- der the label “applications barrier to entry,” this line of reasoning has featured prominently in the Microsoft case. See, e.g., United States v. Microsoft Corp., 253 F.3d 34, 54-56 (D.C. Cir. 2001); Gilbert & Katz, supra, at 28-30. 101. In addition to offering its own content and applications to the customers of its Inter- net service, the monopolist may also “allow” independent producers of these products to offer their products to the customers of its Internet service, as long as they agree to offer their prod- ucts exclusively to these customers. Stated differently, instead of depriving a potential entrant into the market for Internet services of a source of complementary products by driving rival content and application producers from the market, the monopolist could deprive the potential entrant of a source of complementary products by signing exclusive contracts with independ- ent content and application producers. Whether a monopolist could profitably impose such an

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

360 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

customers of rival network providers could simply use the monopolist’s content and applications with the rival’s Internet service.102 Hence, for a particular application or content, this strategy and the “primary good not essential” strategy are mutually exclusive.103

Thus, this theory is only applicable, if (a) an Internet service pro- vider offers proprietary content and applications exclusively to customers of its Internet service,104 and if (b) – potentially due to the exclusion of rivals from its customers – there are not enough remaining independent applications, content or portals available that could be used by customers of rival or newly entering network providers.105 In this case, a new en- trant into the market for Internet services needs to develop (or subsidize the development of) its own content or applications.

One may wonder whether the condition (b) may ever be fulfilled in

exclusivity provision, has been the subject of considerable debate. The Chicago school denied such a possibility, arguing that the other party to the exclusive contract would not agree to con- tracts that made it worse off, e.g., BORK, supra note 39, at 309. More recent research has shown that this argument is incomplete: it does not consider the possibility that the exclusive contract imposes harm on third parties that are not parties to the contract, while not making the contracting parties worse off. In other words, the exclusive contract gives rise to a negative externality on third parties, and due to this externality, signing an exclusive contract is jointly optimal for the contracting parties. For a discussion of this question with pointers to the litera- ture, see, for example, Gilbert & Katz , supra note 100, at 31-33; Michael D. Whinston, Exclu- sivity and Tying in U.S. v. Microsoft: What We Know; and Don’t Know, 15 J. ECON. PERSP. 63, 66-70 (2001). 102. Usually, this theory is applied to cases, where the entrant’s primary good is techni- cally unable to take advantage of the set of applications developed for the monopolist’s pri- mary good. For example, software applications make use of a specific operating system’s ap- plication programming interfaces and therefore run only on this operating system. As a result, customers of the entrant’s operating system are technically unable to use applications devel- oped for the incumbent’s operating system. By contrast, as long as an application complies with the specifications of the Internet protocol, it can run over any physical network that sup- ports the Internet protocol. As a result, applications adhering to that standard can be used by anyone connected to the Internet. Thus, from a technical point of view, the applications offered by the monopolist could be used by customers of a rival network provider as well. Therefore, the entrant’s inability to use the monopolist’s applications and content is not due to technical differences or incompatibility between the Internet services offered by the monopolist and a potential entrant, but results from the monopolist’s business decision to offer its content and applications exclusively to customers of its own Internet service. 103. The strategy described here requires that the monopolist does not offer the content, application or portal to consumers outside its network; by contrast, in the “primary good not essential” strategy, the inability to earn monopoly profits on its sales to consumers outside its network is the reason that leads the monopolist to monopolize the complementary market as well. See supra Part II.B.3.1. 104. The potential anti-competitive implications of such a strategy are explored by, for example, MacKie-Mason, supra note 47, at 23-25; Rubinfeld & Singer, supra note 13, at 313- 16. 105. Alternatively, the monopolist could reach the same result by allowing independent producers of applications, content and portals to offer their products to the customers of its Internet service, if they agree to provide the products exclusively to its customers. See the dis- cussion supra note 101.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 361

the Internet context: after all, there are a number of portals, content and applications that are available to anyone using the Internet today. The condition may be met in emerging markets such as the market for broad- band Internet services, the market for Internet services for mobile phones or in emerging national markets in countries outside the United States. For example, there may be not enough independent applications or con- tent that take advantage of broadband specific characteristics such as high transport speed or broadband’s always on capacity.106 Similarly, there may not be enough independent applications or content that are adapted to the specific limitations associated with using the Internet from mobile phones.107 In a country that just started adopting the Internet, there may not be enough independent applications or content in the na- tional language.

One may also imagine that consumers perceive certain applications and content as indispensable elements of Internet usage. If these applica- tions and content are exclusively available with the incumbent’s Internet service, consumers may not consider an entrant’s Internet service an adequate alternative to the incumbent’s Internet service, unless the en- trant offers a similar set of applications and content itself. In this case, to deter entry to the market for Internet services, the incumbent does not need to drive all existing independent applications, portals and content from the market and offer all affiliated complementary products exclu- sively to customers of its Internet service. It suffices to restrict the exclu- sionary conduct to those applications and content that consumers view as essential. Although there are independent applications and content left that customers of a rival Internet service could use, the entrant will still be forced to enter the market for specific applications and content in or- der to be able to compete in the primary market.108

Second, simultaneously entering the market for Internet services and the market for content or applications must be more difficult or costly than entering the market for Internet services alone. This require- ment is fulfilled as well. Simultaneous entry into both markets is more difficult or costly than entry into the market for Internet services alone if

106. Many broadband customers may simply use broadband Internet services to access narrowband offerings at higher speed. According to McKinsey, “so far, [. . .] faster and better access to the Internet is the sole killer application of broadband.,” Beardsley et al., supra note 67, § “what happens next?” and Exhibit 6. 107. For example, compared to PCs, mobile handsets have small screens, limited key- pads and not a lot of storage. See, e.g., Francis Deprez et al., Portals for All Platforms, in MCKINSEY Q., Issue 1, 2002, at 92. 108. Finally, one may imagine a situation in which the nationwide market for Internet services consists of a collection of local monopolies who all bundle their content, portal and applications exclusively with their Internet service. In this case, a new entrant into the market for Internet services would have to enter the market for content, portals, or applications as well.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

362 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

the two markets exhibit at least one of the four characteristics described above. In the Internet context, all four characteristics are present: first, entry to both markets requires very different capabilities, second, produc- tion in both markets is subject to differing economies of scale, third, suc- cess in the different markets is uncertain, and finally, due to the incum- bent’s exclusionary conduct, the provision of Internet service is subject to indirect network effects with respect to the individual provider’s net- work.

First, developing software applications or interesting content re- quires very different capabilities than operating a network. As a result, a potential entrant to the market for Internet services may not necessarily have the capabilities required for entering the markets for applications or content.109 In addition, most of the cost of entry into those complemen- tary industries consists of the sunk costs of developing the offering that cannot be recovered in case of failure.110 Due to these factors, the need to enter the complementary markets as well considerably increases the risks associated with entry to the primary market. Consequently, an entrant into both markets will most likely be charged higher rates for capital than an entrant to the primary market alone.

Second, the market for Internet services and the markets for com- plementary products are subject to very different economies of scale: for example, McKinsey estimates that assuming an average revenue per user of $ 18.00 to $ 22.50 a year in 2005, a broadband PC portal in Germany would need more than 8 million unique users to break even.111 By con- trast, the economies of building and operating physical networks over which IP services could be provided are much lower.112

Third, although network technology is undergoing rapid innovation, a new entrant into the market for Internet services can take advantage of existing technology and does not have to innovate itself. By contrast, the development of applications and content is subject to considerable uncer- tainty. If a potential entrant to the market for Internet services needs to develop several applications and services in order to be able to compete with the incumbent’s Internet service, the uncertainty associated with each development reduces the likelihood of successful entry to the mar- ket for Internet services.

109. See, e.g., Robert Niewijk et al., Why European ISPs Need Partners, in MCKINSEY Q., Issue 1, 2003, 98. 110. That the costs of capital may increase with the amount of entry costs that are sunk is discussed by W. KIP VISCUSI ET AL., ECONOMICS OF REGULATION AND ANTITRUST 157-58, 161 (3d ed. 2000). 111. Deprez, et al., supra note 107. 112. For example, as of June 30, 2001, the 10 largest providers in the market for broad- band transport services in the United States had between 1,409,000 and 360,000 residential broadband customers, Yoo, supra note 31, at 256 tbl.7 (internal citations omitted).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 363

Fourth, Internet service is subject to indirect network effects:113 the more applications and content are available for users, the more valuable Internet service becomes. At the same time, the development of content and applications is subject to economies of scale.114 As a larger number of users allows application and content developers to spread the fixed costs of development over a larger potential sales base, under free entry to these markets the variety of applications and content will be higher and their cost will be lower, the larger the number of users.

Technically, any application based on the Internet protocol can run over any network that is connected to the public Internet and supports the Internet protocol. As a result, from a technical point of view, the relevant network for indirect network effects is not an individual provider’s net- work, but the global Internet. Thus, technically, Internet service provid- ers compete under conditions of compatibility.

By excluding independent applications from its network and offer- ing its own applications exclusively to its own Internet customers, an Internet service provider can reintroduce indirect network effects with respect to its own network.115 Stated differently, as a result of this strat- egy, the benefits of adding a new user do not accrue to anyone connected to the Internet, but are limited to the customers of the new user’s Internet service provider. Application and content developers now have to decide whether to offer their product to the customers of the Internet service provider with the “closed” network or to the customers of Internet ser- vice providers following an open system strategy. Due to economies of scale in the production of application and content, the developers will base their decision on the size of the different networks.

As a result, an entrant to the market for Internet services will have difficulties attracting application and content developers who write for its network instead of the incumbent’s. Thus, due to the incumbent’s strat- egy, the entrant faces the chicken and egg problem described above: con- sumers will not subscribe to its Internet service in the absence of an at- tractive amount of content and applications; application and content developers will not produce for its network in the absence of an attractive number of users.116

113. E.g., Speta, Handicapping, supra note 5, at 83-84. 114. See supra Part II.B.2.3.b. 115. An Internet service provider could reach the same effect (i.e., reintroduce indirect network effects with respect to its own network) by using proprietary protocols inside its net- work, see, e.g., COMPUTER SCI. & TELECOMM.S BD. & NAT’L RES. COUNCIL, THE INTERNET’S COMING OF AGE 147-49 (2001). An alternative strategy may be the provision of quality of service only within an Internet provider’s network, see, e.g., SHAPIRO & VARIAN, supra note 48, at 187. 116. COMPUTER SCI. & TELECOMM.S BD. & NAT’L RES. COUNCIL, supra note 115, at 147-49, describe a similar situation in the context of provider-specific indirect network effects

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

364 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

Thus, a monopolist provider of Internet services may be able to de- ter entry to the market for Internet services by excluding rival producers of applications, content and portals from the market and offering its own content and applications exclusively to the customers of its own Internet service.117 This strategy may reduce consumers’ valuation of its Internet service, as the exclusion of rival producers of applications, content and portals reduces the variety of complementary products available to cus- tomers of its Internet service. Thus, in deciding whether to employ such a strategy, the monopolist must trade off the loss in Internet service fees against the gains in future monopoly profits.

4. Profitability of Discrimination without Monopolization

In the network neutrality context, researchers commonly focus on the ability and incentive of a network provider to monopolize the market for selected complementary products. The previous sections have fol- lowed this approach. It is based on the implicit assumption that discrimi- nation is only profitable, if the network provider manages to monopolize the complementary market. As the following section shows, this focus may be too narrow: A network provider may have an incentive to dis- criminate against an application even if the provider does not manage to drive it from the market.

Thus, researchers commonly underestimate the likelihood of dis- criminatory behavior by network providers: If discrimination requires the network provider to monopolize the complementary market to be a prof- itable strategy, discrimination will be restricted to those cases where the network provider can expect to drive its competitors from the comple- mentary market. If, however, discrimination is a profitable strategy, even if the network provider does not manage to monopolize the complemen- tary market, it is much more likely to occur.

The following analysis will cover four of the five exceptions out- lined above.118 It is based on the assumption that the exclusion of rivals from the network provider’s Internet service customers increases the number of sales of the network provider’s complementary product. At least some of the network provider’s Internet service customers that would have used a rival’s complementary product in the absence of ex-

due to the use of proprietary protocols inside the network. 117. As highlighted supra note 101, an alternative way of deterring entry would be to sign exclusive contracts with independent producers of applications, content and portals. Such a strategy would have the advantage that the monopolist does not have to bear losses with re- spect to its Internet service fees, as its customers would have access to all existing applica- tions, content and portals. 118. The fifth exception, “monopoly preservation in the primary market,” supra Part II.B.3.2, requires that rival producers of excluded complementary products are driven from the market.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 365

clusion will use the network provider’s offering instead. Thus, by exclud- ing rival producers of applications or content from its network, the net- work provider gains additional sales from its Internet service customers at the expense of its rivals. If the complementary product is subject to economies of scale or network effects and the network provider offers its complementary product to customers nation-wide, the exclusion from the network provider’s Internet service customers may force rivals to operate at an economically less efficient scale or with a smaller network of cus- tomers, putting the rivals at a competitive disadvantage in the rest of the market as well and potentially leading to even more additional sales for the network provider’s complementary product.

Based on this assumption, the analysis will ask, whether a larger number of sales of the network provider’s complementary product in- crease its profits, even if the network provider does not manage to mo- nopolize the complementary market in question.

4.1. More Sales at Market Prices

In a perfectly competitive market subject to constant returns to scale, simply increasing the number of sales at the market price will not increase profits. In such an industry, long-run equilibrium prices equal marginal costs, resulting in zero profit per unit sold. As a result, a firm cannot increase its profits by making additional sales at the market price. Instead, it needs to gain a monopoly position that enables it to raise prices above marginal costs.

Markets for applications, content and portals are different: In these markets, the exclusionary conduct need not result in a monopoly to in- crease the network provider’s profits; it suffices if it results in a larger number of sales.119 This is due to the cost structure underlying the pro- duction of applications and content: the production of these goods is characterized by high fixed costs and very low marginal costs. While the costs of developing the first instance of an application or content may be significant, the costs of producing additional copies may be negligible. Due to the need to cover fixed costs, such products are priced signifi-

119. For an economic model demonstrating this effect in the context of tying, see Patrick DeGraba, Why Lever into a Zero-Profit Industry: Tying, Foreclosure, and Exclusion, 5 J. ECON. & MGMT. STRATEGY 433 (1996). In DeGraba’s model, oligopolists sell a differentiated good (the primary good) and a homogenous good (the complementary good) that are used in fixed proportions to produce the final good. The homogenous good can be produced at con- stant marginal cost by any firm incurring a certain fixed cost. The homogenous market is char- acterized by free-entry Cournot competition. In such a market, the zero-profit price of the good is greater than the marginal cost. As a result, the oligopolist in DeGraba’s model will tie in order to increase the sales of the complementary good. Note that this model does not require the complementary good to be a differentiated good.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

366 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

cantly above marginal costs.120 If the market price is significantly above marginal costs, a firm does

not need to be able to charge monopoly prices to increase its profits: In- stead, making additional sales at the market price may be enough.121 More sales enable the firm to spread the fixed costs of production over more units, resulting in lower average costs per unit and a higher profit margin at the same price. Put differently, once a firm has made enough sales to cover the fixed costs, any additional sale at the market price only adds to the profits. For example, given that gross margins of 80% or 90% are common in computer software,122 any additional sale may lead to a significant increase in profits.

By excluding rival producers of complementary products from its network, the network provider gains additional sales. These additional sales increase the network provider’s profit, even if the excluded rivals are not driven from the complementary market completely.

For example, this fact has important implications for the relevance of the exception “primary good not essential” outlined above.123 Whether a network provider can monopolize the nation-wide complementary market in question by excluding its rivals from access to its Internet ser- vice customers, depends on a variety of factors such as the exact size of economies of scale with respect to the complementary product in ques- tion, the strength of any potential network effects and the size of both the monopolist’s network and the remaining network. Ultimately, the cases in which monopolization is a realistic prospect may not be very common. As monopolization is not necessary to increase the network provider’s profits, however, this restriction does not matter. As long as the exclu- sion of rivals from its Internet service customers enables the network provider to increase the number of sales of its complementary product and the additional profits resulting from more sales at the market price are larger than the costs of exclusion, exclusion will be a profitable strat- egy. Given how often the conditions underlying the “primary good not essential” exception124 are met, this drastically increases the likelihood

120. If the price were equal to marginal costs, firms would not be able to cover their fixed costs and would earn negative profits. In the long run, firms would not operate in such a market. Thus, even if all firms earn zero profit per unit in long-run equilibrium, equilibrium prices are above marginal costs. 121. SHAPIRO & VARIAN, supra note 48, at 161. The importance of market share and number of units sold in knowledge-based products is also described by AFUAH & TUCCI, supra note 49, at 52-54. For an economic model demonstrating this effect in the context of tying, see DeGraba, supra note 119. 122. Katz & Shapiro, supra note 70. 123. See supra Part II.B.3.1. 124. As outlined supra Part II.B.3.1, these conditions are: The network provider has a monopoly in the primary market (i.e., the market for Internet services). The primary good is not essential (i.e., there are uses of the complementary product that do not require the primary

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 367

that exclusion may be a profitable strategy.

4.2. More Outside Revenue

As indicated above,125 a network provider may have an incentive to monopolize the complementary market, if the complementary product is a source of outside revenue that cannot be extracted in the market for Internet services. For reasons outlined above, its outside revenue will be higher if it excludes its rivals and collects the outside revenue directly than if tries to capture some or all of its rivals’ outside revenue by threat- ening exclusion.

This increase in profit, however, is not dependent on a monopoliza- tion of the complementary product market.

Although the network provider’s revenue from outside sources will be highest if it manages to monopolize the market for access to its cus- tomers, increasing the number of customers who access the network pro- vider’s offering may still lead to higher profits than trying to extract the outside revenue from its rivals.

Evidence suggests that even without a monopoly, the relationship between the number of customers and advertising revenue is not a linear one: for example, MacKie-Mason reports that although Lycos had 72 percent as many unique visitors as Yahoo! in September 1999, it re- ceived only 36 percent as much advertising revenue.126 This implies that selling access to one large group of customers as a whole may still yield substantially more revenue than selling access to subgroups of that group separately, even if the seller does not have a monopoly in the market for access to its customers.

In addition, through its billing relationship with customers of its Internet service, the network provider has data on customer demograph- ics that enables it to charge higher advertising fees or commissions for online sales than many of its rivals in the market for Internet content, portals and applications.127 Again, this ability is not dependent on a mo-

good). This condition is met when the Internet service provider offers its complementary prod- uct not only to its Internet service customers, but to customers nation-wide. The complemen- tary market is subject to economies of scale or network effects, a condition that is met in most markets for applications, content or portals. The monopolist has a mechanism at its disposal that enables it to exclude its rivals from access to its primary good customers. In the Internet context, technology that enables the network provider to distinguish between applications run- ning over its network and to control their execution provides the network provider with this capability. 125. See supra Part II.B.2.1. 126. MacKie-Mason, supra note 47, at 13. 127. Even if those rivals require consumers to register before using their product or ser- vice, they have no way to verify the information, unless they require payment; in this case, they can verify the information as part of the billing process, SHAPIRO & VARIAN, supra note 48, at 34-35; MacKie-Mason, supra note 47, at 11.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

368 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

nopoly in the complementary market. A similar argument applies to the variant of this exception described

above.128 In this variant, a network provider excludes Voice over IP (VoIP) providers from access to its Internet Service customers in order to preserve the outside revenue in the form of access charges associated with traditional long-distance calls. Such a strategy will also be profit- able, if the network provider does not manage to exclude the VoIP pro- viders from its customers completely: Access charges are per-call charges set by regulators; the ability to charge them is not dependent on keeping all long-distance customers. Every long-distance call lost to a VoIP provider reduces profits; the more conventional long-distance calls the network provider manages to keep, the higher its profits.

4.3. Monopoly Preservation in the Complementary Market

In the “monopoly preservation in the complementary market” ex- ception outlined above,129 the network provider excludes rival producers of a complementary product from access to its Internet service customers to preserve a legally acquired monopoly in the corresponding comple- mentary market.

In the exception outlined above, the analysis assumes that the mo- nopolist will be able to keep its rivals out of the nation-wide market by excluding them from access to its Internet service customers.

Even if the monopolist’s footprint is not large enough to force its ri- vals to stay out of the market completely, exclusion from a part of the market may put them at a severe competitive disadvantage by forcing them to operate at a less efficient scale or with a smaller network. Com- pared to a world without exclusion, this may slow down the erosion of the network provider’s monopoly in the complementary market, preserv- ing its ability to charge monopoly profits for a longer time. Again, this may make exclusion a profitable strategy, even if the network provider does not manage to keep its rivals out of the market completely.130

C. Network Provider Faces Competition in the Market for Internet Services

Up this point, the analysis was based on the assumption that the network provider is at least a local monopolist in the market for Internet services. This assumption is in line with standard economic thinking on

128. See supra Part II.B.2.2. 129. See supra Part II.B.2.3. 130. Cf. POSNER, supra note 39, at 254 (making a similar argument with respect to the profitability of monopoly preservation through exclusionary conduct in new economy markets, if the monopoly is of intellectual property).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 369

vertical exclusionary conduct in complementary markets: according to economic theory, an economic actor without monopoly power in the primary market will be incapable of excluding competitors in the com- plementary market using vertical practices such as tying, vertical mergers or exclusive dealing. A monopoly in the primary market is therefore con- sidered to be an indispensable precondition for successful monopoliza- tion of the secondary market.131

Given this theory, it is not surprising that most of the literature on vertical exclusionary conduct in complementary product markets focuses on exclusionary conduct by monopolists: after all, the same conduct is unlikely to pose any significant anti-competitive threat, if the firm faces competition in the primary market.132 This theory has also shaped the evaluation of existing firms’ behavior in a complementary market: alle- gations of anti-competitive conduct in a secondary market are often countered by evidence that the accused firm does not have monopoly power in the primary market.133 Alternatively, the analysis of the mo- nopoly case is used as an argument “a maiore ad minus”: if a monopolist in the primary market does not have the ability and incentive to impede competition in the secondary market, it is argued, then a competitive firm’s conduct will pose even less of a threat.134

Based on this line of reasoning, most commentators believe that the threat of discrimination against independent providers of complementary products can be mitigated by competition in the market for Internet ser- vices. Stated differently, it is usually assumed that competition in the market for Internet services will restrict a network operator’s ability and incentive to discriminate against independent content, portals or applica- tions. This assumption forms the basis for two common policy proposals:

131. E.g., id. at 195; Yoo, supra note 31, at 188-91. Similarly, some sort of market power or political power is considered to be a prerequisite for strategies that raise rivals costs, e.g., DENNIS W. CARLTON & JEFFREY M. PERLOFF, MODERN INDUSTRIAL ORGANIZATION 353 (3d ed. 2000). 132. For an important exception to this point, see the literature on the exercise of after- market power by a firm that faces competition in the foremarket. This literature focuses on the question whether primary market competition precludes anti-competitive aftermarket actions. For an analysis of these issues with pointers to the literature, see Jeffrey K. MacKie-Mason & John Metzler, Links between Vertically Related Markets: Kodak, in THE ANTITRUST REVOLUTION: ECONOMICS, COMPETITION, AND POLICY 386 (John E. Kwoka, Jr. & Lawrence J. White eds., 3d ed. 1999). 133. See, e.g., Yoo, supra note 31, at 249-50, 253 in the context of the open access de- bate (“I conclude that the structure of the broadband industry renders it unlikely that such combinations will pose any significant anti-competitive threat. . .”); and Yoo, Beyond Network Neutrality, supra note 4, at 61 in the context of the network neutrality debate (“This suggests that for most of the country, competition should remain sufficiently robust to ameliorate con- cerns of anticompetitive effects.”). 134. E.g., Speta, Vertical Dimension, supra note 5, at 986 (discussing this notion in the context of the open access debate).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

370 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

the first proposal assumes that fostering facilities-based competition (i.e., increased competition between operators of different physical networks) will mitigate a network provider’s ability and incentive to discrimi- nate.135 The second proposal seeks to restore competition at the Internet service provider level by requiring the owners of broadband networks to allow independent Internet service providers to offer their services over these networks. This regulatory response is called “open access,” “multi- ple access” or “forced access,” depending on the point of view of the commentator.136

The following analysis shows that this assumption is not correct: a network provider may have the ability and incentive to exclude rival con- tent, applications or portals from its network, even if it faces limited competition137 in the market for Internet services.138 Apart from increas- ing the number of cases in which unaffiliated providers of complemen- tary products face a real threat of discrimination, this result also implies that neither facilities-based competition nor open access regulation are the appropriate tools to mitigate this threat.139

Three arguments drive this result: First, in the Internet context, the ability to exclude competitors from a complementary market (the mar- kets for applications, content and portals) is not dependent on a monop-

135. See, e.g., Yoo, Mandating Network Neutrality, supra note 4, at 67 (“On the other hand, regulators can adopt a more humble posture about their ability to distinguish anticom- petitive from procompetitive behavior and attempt to resolve the problem by promoting entry by alternative broadband platforms. Once a sufficient number of alternative last mile providers exists, the danger of anticompetitive effects disappears, as any attempt to use an exclusivity arrangement to harm competition will simply induce consumers to obtain their services from another last mile provider”). 136. An example of this line of reasoning can be found in the FCC memorandum and opinion in the AOL Time Warner merger proceeding. AOL Memorandum Opinion & Order, supra note 34, at 6594-95, ¶ 107 (“We believe that if unaffiliated ISPs receive non- discriminatory access to Time Warner cable systems [. . .] the merged firm’s incentives and ability to withhold unaffiliated content from its subscribers will be substantially mitigated.”); see id. at 6596, ¶ 112; Lemley & Lessig, Ex parte, supra note 31. 137. The analysis assumes that the network provider competes with at least one other network provider. See infra note 140 and accompanying text. 138. See also Joseph Farrell, Open Access Arguments: Why Confidence is Misplaced, in Net NEUTRALITY OR NET NEUTERING: SHOULD BROADBAND INTERNET SERVICES BE REGULATED 195 (Thomas M. Lenard & Randoph J. May eds., 2006) (arguing that limited competition may not necessarily remove network providers’ incentives to discriminate). For a similar argument in the context of the debate over censorship by private proxies, see Seth F. Kreimer, Censorship by Proxy: the First Amendment, Internet Intermediaries, and the Prob- lem of the Weakest Link, 155 U. PA. L. REV. 11, 33-36 (2006) (arguing that competition be- tween Internet service providers may not be sufficient to discipline Internet service providers that disable content needlessly based on arguments very similar to the ones advanced above). 139. There may be other reasons that justify these proposals, though. For example, ac- cording to Lemley & Lessig, Ex parte, supra note 31, at 21-25, ¶ 54-65, the reduction in appli- cation-level innovation by independent providers resulting from the threat of discrimination constitutes only one of three arguments in favor of open access.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 371

oly position in the primary market (the market for Internet services). In- stead, the power to exclude is conferred by network technology (Section 1). Second, realizing the benefits of exclusion (i.e., an increase in profits (or, sometimes, a preservation of current profits)) does not require a mo- nopoly position in the primary market. The lack of monopoly in the pri- mary market even increases the network provider’s incentive to increase profits by engaging in exclusionary conduct in the complementary mar- ket, as the network provider cannot simply extract the available monop- oly profit by charging higher prices in the primary market (Section 2). Third, due to various factors such as the existence of switching costs or the ability to use discrimination instead of exclusion, the exclusion of ri- vals will not necessarily cause the network provider’s Internet service customers to switch to another provider, making the costs of exclusion lower than is commonly assumed (Section 3).

The following analysis assumes that the network provider competes with at least one other network provider.140 In addition, the network pro- vider may offer content or applications. A particular application or con- tent may be offered to all consumers (affiliated product) or exclusively to the customers of its own Internet service (proprietary product).141

1. Ability to Exclude

Today, technology is available that enables network providers to distinguish between applications and content running over its network and to control their execution. This technology enables the network pro- vider to exclude selected complementary products from its network or to slow down their execution.

This technology enables the network provider to exclude unaffili- ated providers of complementary products from access to its Internet ser- vice customers, independent of a monopoly in the market for Internet services.

While the exclusionary power of the technology does not reach be- yond the network provider’s network, exclusion from the network pro- vider’s Internet service customers may suffice to drive rival producers of complementary products from the nation-wide market, if there are economies of scale or network effects in the complementary market.142

140. This assumption reflects the reality in the broadband market for residential custom- ers in the US. According to a recent study by the United States Government Accountability Office, the median number of broadband providers available to residential users is two. United States Government Accountability Office, Report to Congressional Committees; Telecommu- nications; Broadband Deployment is Extensive throughout the United States, but It Is Difficult to Assess the Extent of Deployment Gaps in Rural Areas, at 18 (May 2006), available at http://www.gao.gov/new.items/d06426.pdf. 141. See supra Part II.A. 142. See supra Part II.B.3.1.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

372 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

Whether this will happen, depends on the exact size of economies of scale with respect to the complementary product in question, on the strength of any potential network effects and on the nation-wide number of both the monopolist’s Internet service customers and the customers of other network providers.143 Thus, in this context, the ability to drive competitors from the nation-wide complementary market depends on the network provider’s nation-wide market share in the market for Internet services. Again, a monopoly position in this market is not required.

2. Benefits of Exclusion

In a variety of cases, the exclusionary conduct will increase (or pre- serve) the network provider’s profits in the complementary market. As the analysis will show, this increase is not dependent on a monopoly po- sition in the market for Internet services; nor does it require the network provider to gain a monopoly in the complementary market.144 Instead, the lack of monopoly in the primary market constrains the network pro- vider’s ability to extract profits in the market for Internet services, mak- ing the ability to realize profits in the complementary market even more attractive. As a result, there are many more cases in which exclusion may be profitable than is commonly assumed.

In general, by excluding rival producers of a specific complemen- tary product from access to the network provider’s Internet service cus- tomers, the network provider will increase the number of sales of its own complementary product.145

As set out in detail above, the increase in the number of sales will often lead to an increase in profits. In the cases outlined above, the in- crease in profits results from an increase in the number of sales, not from the ability to charge monopoly profits. Thus, to be profitable, the exclu- sionary conduct need not drive rivals from the complementary market completely.

In the cases described above, the network provider had a monopoly in the market for Internet services. As the following analysis will show, however, the increase in profits due to exclusion was not dependent on this monopoly position (Sections 2.1 – 2.3). In addition, it will highlight a variant of the “monopoly preservation in the primary market” excep- tion outlined above: the network provider may exclude selected produc- ers of complementary products from access to its customers to protect its

143. Even if the monopolist’s footprint is not large enough to force its rivals to exit the market completely, exclusion from a part of the market may put them at a severe competitive disadvantage by forcing them to operate at a less efficient scale or with a smaller network. See the analysis supra Part II.B.4.1. 144. See supra Part II.B.4. 145. See supra Part II.B.4.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 373

competitive position in the primary market (Section 2.4).

2.1. More Sales at Market Prices

In the exception “more sales at market prices,”146 the increase in profits resulting from the higher number of sales in the complementary market was driven by the specific cost structure of the markets for appli- cations, content or portals, which are characterized by high fixed costs and low marginal costs. This cost structure is not affected by the exis- tence of market power in the market for Internet services.147

2.2. More Outside Revenue

In the exception “more outside revenue,”148 the increase in profits resulted from the logic of pricing in the markets for advertising. This en- abled the network provider to realize higher outside revenue by selling access to a large group of its Internet service customers directly, instead of letting rival producers of complementary products sell access to smaller groups of customers and extracting the outside revenue from them. Again, a monopoly in the market for Internet services is not re- quired for this relationship to hold.

There is evidence that some Internet service providers (i.e., eco- nomic actors that face competition in the Internet service market) do in fact attempt to reduce the amount of time their customers spend on unaf- filiated content or portal offerings. For example, in the AOL/Time War- ner merger proceeding the FCC found that “[t]he record in this proceed- ing provides some evidence that AOL already seeks to limit its members’ access to unaffiliated content on the World Wide Web. For example, AOL requires that content appearing on AOL web sites have only a lim- ited number of hyperlinks to unaffiliated content.” [References omit- ted]149

In the variant of this exception,150 the network provider was inter- ested in excluding Voice over IP (VoIP) providers from access to its cus- tomers, because it could only charge access charges for long-distance calls placed using the conventional telephone service, not for long-

146. See supra Part II.B.3.1, and Part II.B.4.1. 147. DeGraba’s model, DeGraba, supra note 119, which demonstrates this effect in the context of tying, supports this analysis. In the model, the producer of the primary good has an incentive to tie in order to increase the number of sales of the secondary good, although it competes with another producer in the primary market. Thus, in the model the incentive to ex- clude independent competitors from the secondary market is not dependent on a monopoly position in the primary market. The model is discussed in more detail supra note 119. 148. See supra Part II.B.2.1, and Part II.B.4.2. 149. AOL Memorandum Opinion & Order, supra note 34, at 6594 ¶ 106; id. at 6593-94, 104-06. 150. See supra Part II.B.2.2, and Part II.B.4.2.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

374 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

distance calls using VoIP. Access charges are per-call charges set by regulation; they do not depend on a monopoly in the market for Internet services.

2.3. Monopoly Preservation in the Complementary Market

In the exception “monopoly preservation in the complementary market,”151 the ability to preserve the monopoly in the complementary market depended on various factors such as the exact size of economies of scale with respect to the complementary product in question, on the strength of any potential network effects and on the nation-wide number of both the monopolist’s Internet service customers and the customers of other network providers. A monopoly in the market for Internet services is not required.

2.4. Preserving Competitive Position in the Primary Market

The exclusion of rivals may protect the network provider’s competi- tive position in the market for Internet services, even if it faces competi- tion in this market. Such an incentive may occur, if an Internet transport provider offers proprietary content and applications exclusively to its transport customers. This is a common strategy, as it enables the trans- port provider to relax price competition in the market for Internet ser- vices by differentiating its transport service from rival offerings, to re- duce customer turnover and increase profits by raising switching costs and to make additional profits by selling access to its customers to adver- tisers, content providers or online merchants.152

Independent content and applications that can be used from any provider threaten the success of this strategy:

First, they reduce the differentiation of a provider’s offerings by providing comparable, but independent alternatives.

Second, independent offerings may reduce the switching costs of the network provider’s Internet service customers. Switching costs are the costs a customer incurs when switching to a competitor.153 For ex- ample, when switching from one dial up access provider to another, a consumer must reconfigure his or her Internet access program. When switching from broadband access over cable to DSL, a consumer also needs to buy and install new equipment such as a DSL modem. Switch- ing costs reduce customer turnover: when considering whether to switch

151. See supra Part II.B.2.3, and Part II.B.4.2. 152. See, e.g., MacKie-Mason, supra note 47, at 11. 153. See, e.g., VARIAN, supra note 42, at 603-05 (providing overview of switching costs); see also SHAPIRO & VARIAN, supra note 48, chapters 5-6 (treating switching costs in the context of information goods).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 375

to a competitor, a customer takes his switching costs into account. Switching costs also make demand more inelastic, enabling the seller to raise prices.154

Bundling Internet transport service with proprietary content and ap- plications that are offered exclusively to transport customers is a com- mon way to increase switching costs.155 In this case, consumers loose ac- cess to their old provider’s proprietary content and applications when they switch to another provider. As a result, they have to search for new ones and learn how to use them. If the new provider does not offer com- parable content or applications, not being able to use the old provider’s proprietary content or applications any more is itself a cost of switching. In addition, many proprietary offerings induce their customers to engage in nontransferable database creation and customization. For example, Internet service providers offer provider-specific e-mail addresses that cannot be transferred to another provider;156 to take advantage of ser- vices like stock portfolio tracking, instant messaging or customized news pages, users have to enter nontransferable data as well. When switching providers, customers need to notify relevant parties of their new e-mail addresses or instant messaging IDs and loose their site-specific data.

Independent offerings may reduce the effectiveness of this strategy by reducing customers’ switching costs: as the independent application or content is not tied to a specific provider of Internet services, consum- ers can continue to use it after switching providers. In addition, by creat- ing site-specific data on independent offerings, customers can avoid be- coming locked in to a specific access provider.157

Third, as has been set out above, independent alternatives may also reduce the time customers spend using proprietary offerings, reducing third party revenues such as advertising fees or commissions for online sales.

By excluding independent applications and content that compete with the network provider’s proprietary offerings, the network provider may be able to prevent these problems.

3. Costs of Exclusion

Compared to the monopoly case, the existence of other, competing network providers may increase the costs of exclusionary behavior in the complementary market. Due to a variety of factors such as the existence of switching costs or the ability to use discrimination instead of exclu-

154. E.g., VARIAN, supra note 42, at 604-05; Hausman et al., Residential Demand, supra note 31, at 164. 155. See, e.g., MacKie-Mason, supra note 47, at 11. 156. See, e.g., SHAPIRO & VARIAN, supra note 48, at 109-10. 157. They get locked in to the independent offering, though.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

376 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

sion, the costs of exclusion will still be lower than is commonly assumed. If the network provider is the only supplier of Internet services in a

particular geographic area, consumers have no alternative way of access- ing the excluded application or content. They either subscribe to the pro- vider’s Internet service or do not use Internet services at all. Thus, the costs of the exclusionary behavior are twofold: first, the price of Internet services will be lower due to the reduction in application and content va- riety.158 Second, without being able to use the excluded application or content, some consumers may not value Internet services enough to pay the lower price.159 Given that that the pricing of the service already re- flects the reduced value, the number of lost transport customers will probably not be very high.

If the provider competes with at least one other network provider, consumers who desire access to the excluded application may switch to another provider. As these consumers do not have to forgo Internet ser- vices altogether, the number of lost transport customers will probably be higher than if the excluding network provider does not face competition. Thus, competition increases the costs of exclusionary behavior in the complementary market.160

Several factors may limit the costs of exclusionary behavior in spite of competition in the market for Internet services:

First, if the exclusionary conduct manages to drive the producers of the excluded application or content from the market, switching providers will not enable consumers to get access to the excluded product. As a re- sult, fewer consumers will switch in response to the exclusion.161

Second, switching costs may prevent consumers from changing providers to get access to the excluded application.162 This is the case, if the increased value from being able to use the excluded application is smaller than the costs of switching to another network provider. Thus, the higher switching costs, the lower the number of customers lost to other network providers.163

Third, and potentially most importantly, the network provider may be able to avoid this problem altogether by using discrimination instead

158. See Wu, Network Neutrality, supra note 1, at 153 (discussing the costs of a dis- criminatory pricing scheme that prohibits customers of a network provider’s basic Internet ser- vice from using specific applications). 159. See, e.g., Rubinfeld & Singer, supra note 13, at 310. 160. See, e.g., id. at 310. 161. See, e.g., id. at 312-13. 162. See, e.g., Hausman et al., Residential Demand, supra note 31, at 164; Kreimer, su- pra note 138, at 34-35; NUECHTERLEIN & WEISER, supra note 9, at 156. For a discussion of switching costs in the market for Internet services, see supra notes 153-157 and accompanying text. 163. Switching costs do not protect the network provider from losing business from new customers.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 377

of direct exclusion.164 As today’s network technology provides the abil- ity to control the execution of applications running over the network, a network provider can negatively affect the execution of particular appli- cations. For example, the network provider can slow down the transport of certain applications or the delivery of selected content. If a network provider discriminates against a rival’s complementary product, consum- ers’ use of the rival’s product is less satisfactory than their use of the network provider’s own offering, even if the rival’s product is of higher quality.

Thus, discrimination works indirectly by changing consumers’ per- ception of the quality of a rival’s offering. As consumers are unable to detect the true cause of the lower quality, they may mistakenly attribute it to bad product design and use competing products whose use is more satisfactory. For example, a slow gaming experience may be due to bad application programming, insufficient server capacity at the gaming site or slow Internet transport. Similarly, long waiting times for pages from an online shop could result from bad programming of the underlying da- tabases or insufficient server speed. If customers do not usually experi- ence problems with network speed, they will be inclined to blame the online game or the online shop.

With discrimination, consumers have the option of choosing the ri- val’s product, but prefer the network provider’s product which they per- ceive to be of higher quality. Contrary to direct technical exclusion or ty- ing, they will not feel that their choice has been restricted. As they do not wish to use the rival’s product, the discrimination will neither reduce their valuation of the network provider’s Internet services nor cause them to switch to a competing provider.

Thus, if the network provider discriminates against rival products instead of excluding them directly, competition in the market for Internet services does not increase the costs of the exclusionary conduct.

D. Conclusion

Although a network provider does not generally have an incentive to discriminate against independent providers of content, applications or content, the analysis has highlighted a variety of circumstances under which it may have such an incentive. Such an incentive may not only oc- cur if it has a (local) monopoly in the market for Internet services, but also if it faces competition. Whether the conditions giving rise to such an incentive are present in a real life situation, is an empirical question. In most cases, however, the network provider need not be able to gain a monopoly in the complementary market to make exclusion a profitable

164. See, e.g., Rubinfeld & Singer, supra note 13, at 310, 313.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

378 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

strategy, making the threat of discrimination more relevant than com- monly assumed.

In most cases, the network provider need not exclude all independ- ent developers of complementary products from its network in order to increase its profits. Instead, it will often be profitable to exclude only those complementary products that directly compete with one of its own complementary products. This reduces the costs of exclusion, as the re- duction in complementary goods variety is restricted to those products that are actually excluded.

Due to the specific characteristics of markets for applications and content such as the cost structure of information goods and (sometimes) the existence of network effects, the exclusion of rivals may lead to gains that are significantly higher than in traditional markets. As a result, it is more likely that the gains from exclusion exceed the associated costs, making it more likely that exclusion is a profitable strategy.165

II. IMPACT ON APPLICATION-LEVEL INNOVATION

The previous part has highlighted conditions under which a network provider may have an incentive to exclude independent producers of ap- plications, content or portals from access to its Internet service custom- ers. When these conditions are present, independent producers of com- plementary products face a real threat of discrimination.

The following section analyzes the impact of this threat on innova- tion in the markets for applications, content and portals (“application- level innovation”). It shows that the threat of discrimination reduces the amount of application-level innovation by independent producers of complementary products (Section A). While discrimination increases network providers’ incentives to engage in application-level innovation, this increase cannot offset the reduction in innovation by independent producers (Section B). Thus, the threat of discrimination reduces the amount of application-level innovation.

A. Incentives of Independent Producers of Complementary Products

In the absence of network neutrality regulation, the threat of dis- crimination reduces the amount of application-level innovation by inde- pendent producers of complementary products in three ways.

First, when the conditions for profitable exclusion outlined above are present in a particular complementary market, a network provider will discriminate against rivals in this market. As indicated above, dis-

165. Cf. POSNER, supra note 39, at 254 (discussing the profitability of monopoly preser- vation through exclusionary conduct in new economy markets, if the monopoly is of intellec- tual property).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 379

crimination will reduce their profits.166 A potential innovator bases its decision to innovate on the expected costs and benefits of realizing the innovation. Facing the threat of discrimination, potential innovators in affected markets will expect lower profits. Thus, the threat of discrimina- tion reduces their incentives to innovate.

Second, the profitability of exclusion depends on a large number of factors that may not be common knowledge for all market participants. As a result, an economic actor with an idea for a complementary product may not be able to decide whether the network provider will have an in- centive to exclude the final product from the market.167 As a result, po- tential innovators face a significant uncertainty with respect to their fu- ture competitive environment. This uncertainty may reduce a developer’s incentive to innovate, even if the factual conditions for profitable exclu- sion are not present.

Third, the above analysis suggests that independent producers of complementary products need not be concerned about exclusion, if the network provider does not currently offer a competing product. This seems to imply that innovation will only be harmed where the network provider is already vertically integrated into one or more complementary markets. Economic theory shows that this is not correct: Even if the net- work provider does not currently offer a competing product, it may be tempted to imitate the entrant, exclude the entrant from its network and exploit the complementary market itself, once the entrant starts to make significant profits.

Economic models show that in the presence of demand uncertainty in a complementary market, a primary good monopolist with a selling advantage in this market may have an incentive to let an independent producer enter the complementary market first to let him “test the wa- ters.”168 If the level of demand turns out to be large enough once the de- mand uncertainty is resolved, the primary good monopolist enters the

166. The exclusionary conduct hurts independent producers of excluded complementary products in several ways: first, they are excluded from the part of the complementary market that consists of the network provider’s Internet service customers. As a result, they are unable to make any sales in that market. In addition, due to economies of scale and, potentially, net- work effects in the production of their products, the exclusion from a part of the market may put them at a competitive disadvantage in the rest of the market as well. In the worst case, they may be forced to exit the complementary market completely. If they had made at least some sales to the network provider’s Internet service customers in the absence of the exclusionary conduct, the exclusion will reduce their profits. 167. Similarly, the network provider may fail to assess the situation correctly and dis- criminate against or exclude an independent provider of complementary products, even if none of the conditions under which this conduct would be profitable apply. Farrell & Weiser, supra note 13, at 114-17 (calling this problem “incompetent incumbents” and include it in their list of exceptions to their version of the “one monopoly rent” argument). 168. David A. Miller, Invention under Uncertainty and the Threat of Ex Post Entry, (Aug. 24, 2006), http://ssrn.com/abstract=319180.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

380 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

market as well and uses its selling advantage to make most of the sales. Foreseeing this course of events, the independent producer refrains from entering the market. As a result, nobody enters the complementary mar- ket; there is a region of foregone invention where privately and socially beneficial innovations are not realized.

For this situation to occur, three conditions must be realized: First, there must be demand uncertainty in the complementary market. Second, in the presence of demand uncertainty, entry to the complementary mar- ket is attractive for the independent producer, but not for the primary good monopolist (e.g., due to cost heterogeneity). Third, the primary good monopolist has a selling advantage in the complementary market.

In the Internet context, these conditions will often be met: First, in markets for new applications or content, there is usually a considerable demand uncertainty. Second, the economics and business strategy litera- ture highlights a variety of reasons, why an incumbent network provider may not have an incentive to enter a complementary market for a new product in the presence of demand uncertainty, while an independent producer may have such an incentive. For example, start-ups often have lower entry costs than an incumbent due to the different cost structure of incumbents and new entrants.169 In addition, while a small level of de- mand may meet the growth needs of a small company, a large incumbent will need much higher levels of demand to meet its growth needs.170 Similarly, even if the level of demand is too uncertain for the network provider to justify innovation, users may find it attractive to innovate to meet their own application needs.171 Third, the ability to technically ex- clude a rival producer of complementary products from its network pro- vides the network provider with a huge selling advantage in the comple- mentary market.

Thus, the number of markets in which independent developers’ in- centives to innovate are reduced will be larger than implied by the excep- tions outlined above.

B. Incentives of Network Providers

As the previous section has shown, the threat of discrimination re- duces independent producers’ incentives to innovate in the markets for applications, content or portals. This reduction is only relevant, if it is not offset by a corresponding increase in network providers’ incentives to

169. E.g., CLAYTON M. CHRISTENSEN, THE INNOVATOR´S DILEMMA; WHEN NEW TECHNOLOGIES CAUSE GREAT FIRMS TO FAIL 132 (rev., updated ed., Harper Business 2000) (1997). 170. E.g., id. at 128-30. 171. See van Schewick, supra note 32, at 329-42 (providing pointers to the relevant lit- erature).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 381

innovate in these markets. While network providers’ incentives to inno- vate at this level do rise due to the increase in profit under discrimina- tion, this increase in a few network providers’ incentives to innovate cannot compensate for the reduction in innovation by independent pro- ducers.

There are three reasons for this: First, the reduction in potential in- novators results in less diverse approaches to innovation, with negative consequences for the amount and quality of innovation. Second, with re- spect to particular innovations, economic actors other than the network providers may have an incentive to innovate, while the network provid- ers may lack such an incentive. This further reduces the amount of inno- vation. Third, there are specific benefits associated with specific types of independent innovators which a network provider cannot replicate.

First, while there are a large number of (potential) independent pro- ducers of complementary products, there are only a few network provid- ers. Thus, by reducing the innovation incentives of a large number of in- dependent developers, the threat of discrimination ultimately reduces the number of innovators at the application-level. In the presence of techno- logical uncertainty, market uncertainty or consumer heterogeneity, this reduction negatively affects the amount and potentially the quality of ap- plication-level innovation.

Human beings and, consequently, the firms for which they work have different experiences, capabilities and organizations, a fact that is stressed by research in evolutionary economics and management strat- egy. Due to these differences, economic actors may react very differently when exposed to the same situation. The impact of these differences rises with technological uncertainty, market uncertainty or consumer hetero- geneity. Under these conditions, an increase in the number of potential innovators will result in a more diverse set of approaches to innova- tion,172 and a more diverse set of approaches will be socially benefi- cial:173 It guarantees a more complete search of the space of potential complementary products and decreases the probability that beneficial uses of the platform remain undetected. It increases the expected quality of the resulting products and may increase the amount of heterogeneous consumer needs that are served.

Second, research in economics and management strategy has identi- fied systematic differences in the nature and direction of innovative ac- tivity between different types of innovators. In particular, due to differ- ences in history, economic position and capabilities, the same innovation may be attractive to one type of innovator, but not to another. This re-

172. See id. at 299-305 with pointers to the relevant literature. 173. See the discussion in id. at 305-10 with pointers to the relevant literature.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

382 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

search suggests that there are a large number of cases in which economic actors other than the network provider may have an incentive to realize an innovative idea, while the network provider may lack such an incen- tive. For example, this has been shown for new entrants to a complemen- tary market, for venture-capital backed firms and for users.174 When in- dependent producers loose their incentive to innovate, this innovation will be lost.

Third, there are specific benefits associated with specific types of independent innovators which a network provider cannot replicate. For example, research has shown that the participation of firms backed by venture capitalists may increase the amount and the quality of innovation. Enabling users to innovate, may leave less customer needs unserved. In addition, users often make their innovation freely available to others; as a result, such innovations will reach a higher level of diffusion than a similar innovation of comparable quality that is produced by a network provider which sells the innovation to make a profit.175

In the context of the Internet, technological and market uncertainty as well as user heterogeneity are high,176 suggesting that the reduction in independent producers’ incentives to innovate will have the detrimental impact on application-level innovation outlined above.

III. IMPACT ON SOCIAL WELFARE

Network neutrality rules prevent network providers from discrimi- nating against independent producers of complementary products or ex- cluding them from their network. In the absence of network neutrality regulation, there is a real threat of discrimination (see Part II). Regula- tory intervention to remove this threat is only justified, if the social bene- fits of regulatory intervention are larger than the costs.

As Part III has shown, network neutrality regulation increases the

174. Id. at 311-24 (new entrants), 324-29 (venture capital backed firms), 329-42 (users), based on a discussion of the relevant literature. 175. Id. at 337-42, based on a discussion of the relevant literature. While it is difficult to quantify these benefits, there are indications that they may be significant. For example, surveys indicate that today’s standard commercial products may on average leave between 46% and 54% of customer needs unserved. See Nikolaus Franke & Eric von Hippel, Satisfying Hetero- geneous User Needs Via Innovation Toolkits: The Case of Apache Security Software, 32 RES. POL’Y 1199, 1201-02 (2003). 176. Both network technology as well as technologies for the development of applications are still evolving, creating considerable technological uncertainty. A large number of useful applications are still waiting to be identified; in these areas, market uncertainty is high. The more people and businesses get connected to the Internet, the higher the heterogeneity of Internet users will become. Ultimately, the heterogeneity of Internet users will mirror the heterogeneity of society. As a result, the heterogeneity of user needs is bound to be increasing, not decreasing.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 383

amount of application-level innovation. The increase is only relevant, if it is socially beneficial (Section A). On the cost side, network neutrality rules reduce network providers’ incentives to innovate at the network level and to deploy network infrastructure (Section B.1). While regula- tory intervention has its own costs, these are not covered in detail (Sec- tion B.2). When deciding whether to introduce network neutrality regula- tion, regulators must trade-off the benefits against the costs (Section C).

The analysis shows that the increase in application-level innovation is socially beneficial and that these benefits are more important than the costs.

A. Benefits

Network neutrality rules increase the amount of application-level innovation. This increase is only relevant to public policy, if it increases social welfare. This question can be approached in several ways.

First, one may ask whether the amount of innovation is generally lower than the social optimum. In this case, an increase in the amount of innovation would be socially beneficial.

In dealing with such questions, economists often note that the link between innovation and social welfare is theoretically ambiguous:177 on the one hand, some economic models highlight the possibility that in their desire to capture the rents from innovation, firms may increase the level of investment in research and development above the socially efficient amount.178 On the other hand, the existence of uncompensated spillovers and other factors such as the inability of innovators to perfectly appropriate the increase in consumer surplus lead to the theoretical prediction that firms will not be able to completely appropriate the social gains from innovation, leading them to invest less than the socially optimal amount in innovation.179

A closer look at the underlying models indicates that under conditions of uncertainty this ambiguity may disappear, leading to the insight that the amount of innovation is usually too low, which makes any increase in innovation socially beneficial. In models where firms invest more than the socially efficient amount in innovation, the wedge

177. See, e.g., Jennifer F. Reinganum, The Timing of Innovation: Research, Develop- ment, and Diffusion, in 1 HANDBOOK OF INDUSTRIAL ORGANIZATION 849 (Richard L. Schmalensee & Robert D. Willig eds., 1st ed. 1989); JEAN TIROLE, THE THEORY OF INDUSTRIAL ORGANIZATION 399-400 (1988); Michael L. Katz, Intellectual Property Rights and Antitrust Policy: Four Principles for a Complex World, 1 J. ON TELECOMM. & HIGH TECH. L. 325, 337 sec.C (2002). 178. For an overview of this literature, see Reinganum, supra note 177. For a particular example of such a model, see Partha Dasgupta & Eric Maskin, The Simple Economics of Re- search Portfolios, 97 ECON. J. 581 (1987). 179. See, e.g., TIROLE, supra note 177, at 399-400.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

384 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

between private and social benefits from innovation results from the argument that society does not care which firm is ultimately successful, whereas each individual firm wants to be the winner.180 Thus, these models are based on the implicit assumption that similar approaches by different firms constitute a wasteful duplication of efforts that should better be avoided. As indicated above, such an assumption neglects differences in firm heterogeneity. Once firm heterogeneity is taken into account, having different firms approach a particular problem will often be socially beneficial.

These theoretical insights are supported by empirical studies. They indicate that there is indeed too little innovation, because private firms are typically unable to appropriate all social gains from the innovation.181

Second, one may ask whether in the specific case under analysis there is likely to be less innovation than the socially optimal amount. Innovation in platform products182 and complementary products is subject to two types of externalities that are likely to reduce the amount of innovation below the social optimum:183 while the first operates vertically between the platform product and each complementary product, the second externality operates horizontally between different complementary products.

Due to the complementarity between the platform product and complementary products, innovation in complementary products usually increases demand for the platform product and vice versa. If the platform product and the complementary product are developed by different economic actors, the innovator in a complementary component does not appropriate the positive effect on the platform product, and vice versa.184

Innovation in one complementary product usually increases demand for the platform product, which may in turn positively affect demand for other complementary products. If different economic actors pursue innovation in the different components, each actor does not appropriate the positive effect on the other components. As a result, each actor’s incentives to innovate will be lower than the social optimum.

A common solution to the problems caused by such externalities is

180. See, e.g., Dasgupta & Maskin, supra note 178, at 584-85. 181. See, e.g., Edwin Mansfield et al., Social and Private Rates of Return from Industrial Innovations, 91 Q. J. ECON. 221 (1977); Charles I. Jones & John C. Williams, Measuring the Social Return to R&D, 113 Q.J. ECON. 1119 (1998). 182. A platform product is a product that may be used with a large number of comple- mentary products. See, e.g., Douglas G. Lichtman, Property Rights in Emerging Platform Technologies, 29 J. LEGAL STUD. 615 (2000). 183. This observation is made in two different contexts by Timothy F. Bresnahan & Manuel Trajtenberg, General Purpose Technologies ‘Engines of Growth,?’ 65 J. ECONOMETRICS 83 (1995) and Lichtman, supra note 182. 184. See, e.g., Bresnahan & Trajtenberg, supra note 183, at 94; Farrell & Katz, supra note 41, at 414 and appendix.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 385

integration by all affected parties. The integrated entity internalizes the externalities and has therefore higher incentives to innovate.185 In the current context, this is not a feasible solution: no single economic actor will be able to identify and realize all beneficial uses of the Internet.186

Finally, any assessment of the benefits of additional application- level innovation needs to take account of the character of the Internet as a general purpose technology.187

As a general purpose technology, the Internet has the potential to significantly increase economic growth.188 General-purpose technologies offer a generic functionality that can potentially be applied in a large number of sectors within the economy. As the use of a general-purpose technology spreads throughout the economy and increases productivity in the sectors in which it is applied, the promises for economic growth that this technology holds materialize. At the same time, new applications trigger new advances in the general-purpose technology itself; these advances may in turn spawn the adoption of the general- purpose technology in additional sectors of the economy or may lead to new or improved applications in sectors that already use the technology. Thus, the adoption of general-purpose technologies exhibits increasing returns to scale, leading to potentially enormous increases in economic growth.189

As the positive impact of a general purpose technology stems pri- marily from the productivity increases resulting from its adoption in more and more sectors of the economy, the existence of a general- purpose technology is not sufficient to positively impact economic growth. Instead, the rate at which a general purpose technology affects

185. See Farrell & Katz, supra note 41(discussing some important refinements to this statement). As Farrell & Katz demonstrate, integration between two firms that each are the sole supplier of a component that is complementary with the other does not necessarily in- crease the incentives to invest in socially valuable research and development. (See id. at ap- pendix). In addition, they show that integration between a monopoly supplier of one compo- nent with one of several suppliers of a complementary component may inefficiently lower independent suppliers’ incentives to innovate. 186. See, e.g., Timothy F. Bresnahan & Shane Greenstein, The Economic Contribution of Information Technology: Towards Comparative and User Studies, 11 J. EVOLUTIONARY ECON. 95, 98 (2001); Lichtman, supra note 182. 187. See van Schewick, supra note 32, at 346-49 (providing a detailed exposition of the argument in the text with pointers to the literature). 188. On general-purpose technologies, see, e.g., Bresnahan & Trajtenberg, supra note 183; Bresnahan & Greenstein, supra note 186; and the collection of papers in GENERAL PURPOSE TECHNOLOGIES AND ECONOMIC GROWTH (Elhanan Helpman ed., 1998) [hereinafter GENERAL PURPOSE TECHNOLOGIES]. On the Internet as a general-purpose technology, see, e.g., Richard G. Harris, The Internet as a GPT. Factor Market Implications, in GENERAL PURPOSE TECHNOLOGIES, supra, at 145. 189. E.g., Bresnahan & Trajtenberg, supra note 183; Elhanan Helpman & Manuel Tra- jtenberg, A Time to Sow and a Time to Reap; Growth Based on General Purpose Technolo- gies, in GENERAL PURPOSE TECHNOLOGIES, supra note 188, at 55.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

386 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

economic growth depends on the rate of co-invention190 (i.e., the rate at which potential uses of the technology are identified and realized).

With respect to the Internet, this analysis implies that identifying potential uses for the Internet and developing the corresponding applica- tions is the prerequisite for realizing the enormous growth potential in- herent in the Internet as a general-purpose technology.191

As a result, measures that reduce the amount of application-level innovation have the potential to significantly harm social welfare by sig- nificantly limiting economic growth.

B. Costs

On the cost side, network neutrality rules reduce network providers’ incentives to innovate at the network level and to deploy network infra- structure (Section 1). Regulatory intervention also creates its own costs (Section 2); however, these are not covered in detail.

1. Impact on Incentives at the Network Level

As highlighted in Part II, there is a variety of cases in which dis- crimination increases (or preserves) network providers’ profits. As net- work neutrality regulation prevents network providers from realizing these profits, network neutrality regulation reduces their profits. Due to the complementarity between applications, content and portals on the one hand and Internet services on the other hand, this reduction in profits also affects network providers’ incentives to innovate at the network level and to deploy network infrastructure.192

190. The term “co-invention” denotes the innovative activity associated with identifying and realizing potential uses of the general purpose-technology in particular sectors of the economy, e.g., Bresnahan & Trajtenberg, supra note 183, at 86-88; Bresnahan & Greenstein, supra note 186, at 95-97. 191. See ROBERT E. LITAN & ALICE M. RIVLIN, BEYOND THE DOT.COMS; THE ECONOMIC PROMISE OF THE INTERNET 104-07 (2001) (making a similar observation). 192. See THIERER, supra note 7, at 17-19; OWEN & ROSSTON, supra note 4, at 24-25. See also Yoo, Beyond Network Neutrality, supra note 4, at 27-37, 48-53 (arguing that network neutrality may increase concentration in the market for last-mile broadband access, based on a broader definition of network neutrality that includes mandating interconnection, non- discrimination, rate regulation and the adoption of standardized protocol interfaces such as TCP/IP). As Yoo’s argument is based on the negative impact of measures such as mandating the adoption of standardized interfaces, which are not included in the definition of network neutrality used here, his arguments do not apply to the analysis of this paper. For a discussion of the differences in the usage of the term network neutrality, see supra notes 9-11 and accom- panying text. But see Frischmann & van Schewick, supra note 1 (offering a critical reply to Yoo’s argument).

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 387

2. Costs of Regulation

The costs of network neutrality regulation depend on the chosen form of implementation. While the costs of network neutrality regulation are not the focus of this article, existing literature suggests that the costs of regulation itself will not be significant. In particular, they will be sig- nificantly lower than the costs associated with implementing and over- seeing an open access regime.193

C. Trade-Off

The social benefits and costs outlined above suggest that the intro- duction of network neutrality regulation requires a trade-off: On the one hand, network neutrality regulation increases the amount of application- level innovation, which is critically important for economic growth. On the other hand, it decreases network providers’ incentives to innovate at the network level and to deploy network infrastructure. The following section analyzes the two trade-offs in turn.

1. Application-Level Innovation vs. Innovation at the Network Level

Research on information-technology based general-purpose tech- nologies suggests that increasing co-invention194 is more important than increasing innovation in the general-purpose technology itself. Applied to the Internet, this implies that increasing application-level innovation is relatively more important than increasing innovation at the network level.

In information technology-based general-purpose technologies the incentives to invest in advancing the general-purpose technology itself seem to be higher than the incentives to invest in co-invention,195 making it relatively more important to foster co-invention. This difference is attributed to two factors: first, the science and engineering base of hardware technologies is more developed than the science base of software engineering and of finding attractive business uses. Second, due to their generality, general-purpose technologies have larger markets than the individual applications; after all, while not all users of a general-

193. See, e.g., Weiser, supra note 1, at 79-80. 194. The term “co-invention” denotes the innovative activity associated with identifying and realizing potential uses of the general purpose-technology in particular sectors of the economy. E.g., Bresnahan & Trajtenberg, supra note 183, at 86-88; Bresnahan & Greenstein, supra note 186, at 95-97. 195. Timothy F. Bresnahan, The Changing Structure of Innovation in Computing: Sources of and Threats to the Dominant U.S. Position 10, (July 21, 1998), http://www.stanford.edu/~tbres/research/step.pdf.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

388 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

purpose technology need all applications, all users need the general- purpose technology.

These factors are also present in the context of the Internet, making it reasonable to assume that the imbalance between incentives to innovate found in information-based general-purpose technologies in general also exists in the context of the Internet: Network engineering has a more developed science base than the identification of uses and software engineering. Due to the generality of the networking infrastructure, the market for network technology itself is larger than the market for individual applications.

Thus, compared to the incentives to innovate at the application- level, incentives to innovate at the network level are higher. At the same time, application-level innovation is the main determinant of economic growth. This suggests that increasing the amount of application-level innovation is relatively more important than increasing innovation at the network level.

2. Application-Level Innovation vs. Deployment of Network Infrastructure

As indicated above, network neutrality regulation reduces network providers’ profits. This reduction in profits will also affect their incentive to deploy network infrastructure. This causal relationship, however, does not say anything about the degree to which these incentives are reduced.

Thus, in determining the appropriate trade-off between infrastruc- ture deployment and application-level innovation, two questions must be answered: First, will the reduction in profits reduce the incentive to de- ploy infrastructure below the necessary level? Second, even if this is the case, is allowing network providers to discriminate the appropriate solu- tion to this problem?

First, it is an open question, whether network neutrality regulation will reduce incentives to deploy network infrastructure below the neces- sary level. Not surprisingly, network providers and their industry organi- zations have claimed that this is the case. There are several reasons to doubt this assessment, though: Network neutrality regulation does not forbid network providers to vertically integrate into complementary mar- kets;196 it only bans them from using discrimination to increase their sales at the expense of rivals. Thus, it does not prevent network providers from making profit in complementary markets; it just takes away the ad- ditional profits that could be realized due to discrimination.197 It also

196. E.g., Wu, Broadband Debate, supra note 1, at 89. 197. Whether and, if yes, what form of price discrimination should be forbidden under network neutrality regulation, is still an open question. See supra note 9.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 389

does not prevent them from making profit in the market for Internet ser- vices. As a result, the remaining profit may still be sufficient to motivate them to deploy the necessary infrastructure. Moreover, new wireless technologies may ameliorate the problem by further reducing the costs of broadband infrastructure. Thus, it still needs to be proven that the reduc- tion in profits caused by network neutrality regulation suffices to reduce network providers’ incentives to deploy infrastructure so severely that it becomes relevant for public policy.

Second, even if network providers’ incentives are too low to guar- antee the necessary deployment of broadband infrastructure under net- work neutrality regulation, this does not necessarily imply that network providers should be allowed to discriminate.198 As Michael Katz has put it, “In the antitrust – if not regulatory – context [. . .] U.S. policy rejects the notion that the otherwise illegal maintenance or acquisition of mo- nopoly power in a market can be justified by ‘good’ use of the monopoly profits in that market or another one.”199 Following this line of reason- ing, instead of allowing discrimination, regulators should contemplate other ways of ensuring a sufficient deployment of network infrastructure, if necessary. For example, in light of the severe consequences of stifling application-level innovation for economic growth, subsidizing the de- ployment of broadband infrastructure may be preferable to allowing net- work providers to discriminate.

Thus, in trading off application-level innovation against infrastruc- ture deployment, it seems reasonable to opt for fostering application- level innovation in order to realize the enormous growth potential inher- ent in Internet technology, and to contemplate other ways of ensuring a sufficient deployment of network infrastructure, if necessary.

198. But see Yoo, Beyond Network Neutrality, supra note 4, Part II. Yoo argues that by focusing on competition in the market for applications and content, network neutrality propo- nents are focusing on the wrong policy problem. According to him, policy makers should fo- cus on increasing competition in the market for last-mile broadband access, which is less com- petitive than the markets for applications and content, id. at Part II. In line with this assumption, he mainly rejects network neutrality proposals based on their negative impact on competition in last-mile broadband access. Id. at 27-37, 48-53. Apart from neglecting the dif- ferent impact of innovation in these markets on economic growth, and accompanying text, this analysis fails to take account of the possibility to stimulate competition in the market for last- mile broadband access through other means. See supra notes 187-191, 194-195. In addition, his arguments about the negative impact of network neutrality on competition in the market for last-mile broadband access are based on a much broader definition of network neutrality than the one advocated by network neutrality proponents and used in this paper; as a result, his analysis does not carry over to the case of “pure” non-discrimination rules discussed here. See also supra notes 9-11 and accompanying text; see supra note 192 and accompanying text. For a critical appraisal of Yoo’s work on network neutrality, see Frischmann & van Schewick, su- pra note 1; see Herman, supra note 1. 199. Katz, supra note 177, at 340.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

390 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

CONCLUSION

This paper advances the debate over network neutrality by provid- ing an economic framework within which calls for network neutrality regulation can be analyzed.

The analysis shows that calls for network neutrality regulation are justified: In the absence of network neutrality regulation, there is a real threat that network providers will discriminate against independent pro- ducers of applications, content or portals or exclude them from their net- work. This threat reduces the amount of innovation in the markets for applications, content and portals at significant costs to society.

While network neutrality rules remove this threat, they are not without costs: Apart from creating the costs of regulation itself, network neutrality rules reduce network providers’ incentives to innovate at the network level and to deploy network infrastructure. Thus, regulators face a trade-off. As the paper shows, due to the potentially enormous benefits of application-level innovation for economic growth, increasing the amount of application-level innovation through network neutrality regu- lation is more important than the costs associated with it.

Before network neutrality regulation can be drafted, however, more research is needed. In particular, the open questions surrounding the scope of network neutrality regulation need to be resolved. In addition, the best way of implementing network neutrality rules still needs to be identified.200

The paper also contributes to the debate over “open access” and “facilities-based competition.” As has been set out above, the proposals for “facilities-based competition” and “open access” are based on the as- sumption that competition in the market for Internet services will miti- gate a network operator’s ability and incentive to discriminate against or exclude independent portals, content and applications. The analysis has highlighted a variety of circumstances under which a network provider may have the ability and incentive to discriminate against unaffiliated producers of complementary products or exclude them from its network, even if it faces competition in the market for Internet services. Thus, nei- ther increased facilities-based competition nor open access regulation are the appropriate tools to mitigate the threat of discrimination.

Finally, the paper shows that our intuitions regarding the profitabil- ity of exclusionary conduct that have been shaped by antitrust analysis of markets for conventional goods may be misleading in markets such as the markets for applications, portals and content that are characterized by high fixed costs, low marginal costs and, potentially, network effects, in

200. For an overview of open issues in these areas, see supra note 9 and accompanying text.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

2007] NETWORK NEUTRALITY REGULATION 391

particular if the exclusionary conduct is based technological means.

0502_006_VANSCHEWICK 2/19/2007 10:24:25 AM

392 J. ON TELECOMM. & HIGH TECH. L. [Vol. 5

  • van Schewick network neutrality cover SSRN published version
  • 0502_006_vanschewick_rerun