1 FOR WHOM THE NET TOLLS: A TWO SIDED MARKET ANALYSIS AND PUBLIC POLICY IMPLICATIONS FOR THE NET NEUTRALITY DEBATE By HONG GUO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLME NT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 Hong Guo
3 To my parents Xiaoyan Guo and Changjie Guo, for a lifet ime of love, trust, and support
4 ACKNOW LEDGMENTS First of all, I am most indebted to my dissertation advisors Dr. Kenny Cheng and Dr. Shubho Bandyopadhyay for directing me to this exciting research topic, guiding me through the hard exploration process, and encouraging me all along. Without the ir support this dissertation would not have been possible. I am sincerely grateful to my academic role models Dr. Gary Koehler and Dr. David Sappington for their enlightening suggestions and generous guidance. I would also like to thank Dr. Asoo Vakharia a nd my other committee members Dr. Haldun Aytug and Dr. Jane Feng for providing insightful comments and other academic support. I also thank Dr. Praveen Pathak, Dr. Anand Paul, Dr. Selwyn Piramuthu, Dr. Janice Carrillo, and Dr. Aydin Alptekinoglu for their day to day encouragement and continued support. Last but not least, I thank all my fellow PhD students, especially Sean Marston and Yipeng Liu, for sharing with me their ideas and making my four years in Gainesville an enjoyable experience.
5 TABLE OF CONT ENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 TABLE OF CONTENTS ................................ ................................ ................................ ................. 5 LIST OF TABLES ................................ ................................ ................................ ........................... 9 LIST OF FIGURES ................................ ................................ ................................ ....................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 13 What is Net Neutrality? ................................ ................................ ................................ .......... 13 The Two Sides of the Debate ................................ ................................ ................................ .. 15 Structure of the Dissertation ................................ ................................ ................................ ... 18 2 A TECHNOLOGICAL ECO NOMIC FRAMEWORK ................................ ......................... 20 The Technological Perspective ................................ ................................ ............................... 20 The Economic Perspective ................................ ................................ ................................ ..... 21 Literature Review ................................ ................................ ................................ ................... 22 3 CONTENT PROVIDER DISCRIMINATION ................................ ................................ ...... 27 The Model ................................ ................................ ................................ ............................... 27 Net Neutrality ................................ ................................ ................................ ......................... 33 Content Decisions for Consumers ................................ ................................ ................... 33 Pricing Decisions for the BSP ................................ ................................ ......................... 34 No Net Neutrality ................................ ................................ ................................ ................... 34 Content Decisions for Consumers ................................ ................................ ................... 34 Delivery Service Decisions for Content Providers ................................ .......................... 36 Pricing Decisions for the BSP ................................ ................................ ......................... 39 Winners and Losers Comparison Between NN and NNN ................................ ................... 44 Capacity Expansion Decision Incentive to Expand Infrastructure Capacity? ................................ ................................ ..... 46 Decisions for Both Content Providers and Consumers in the Capacity Expansion Problem ................................ ................................ ................................ ........................ 47 Pricing and Capacity Expansion Decisions for the BSP in the Capacity Expansion Problem ................................ ................................ ................................ ........................ 47 4 CONTENT PROVIDER DISCRIMINATION AND MARKET COVERAGE .................... 59 The Model ................................ ................................ ................................ ............................... 60 Net Neutrality ................................ ................................ ................................ ......................... 61
6 No Net Neutrality ................................ ................................ ................................ ................... 62 Computational Analyses ................................ ................................ ................................ ......... 68 Solution Procedure ................................ ................................ ................................ .......... 68 The Baseline Model ................................ ................................ ................................ ......... 71 Research Findings ................................ ................................ ................................ ................... 74 The Fate of the Consumer s Under NNN ................................ ................................ ......... 75 ................................ ................................ .......... 75 The Effect of NNN on Market Coverage ................................ ................................ ........ 76 The Effect on Consumer Surplus ................................ ................................ .................... 77 The Effect on Social Welfare ................................ ................................ .......................... 78 Can NNN Reduce Competition among Content Providers? The Issue of Innovation at the Edge ................................ ................................ ................................ 79 The Possibility of Free Internet Access to Consumers ................................ .................... 80 5 CONTENT P ROVIDER DISCRIMINATION AND VERTICAL INTEGRATION ............ 82 The Model ................................ ................................ ................................ ............................... 83 Content Providers ................................ ................................ ................................ ............ 84 Vertically Integrated BSP ................................ ................................ ................................ 84 Consumers ................................ ................................ ................................ ....................... 85 Net Neutrality ................................ ................................ ................................ ......................... 87 No Net Neutrality ................................ ................................ ................................ ................... 87 ................................ ................................ ............... 96 The Welfare Effect of Net Neutrality ................................ ................................ ..................... 98 Welfare Comparison with Vertical Integration (VI) and No Vertical Integration (NVI) ..... 100 6 USER DISCRIMINATION ................................ ................................ ................................ .. 110 The Model ................................ ................................ ................................ ............................. 114 Net Neutrality ................................ ................................ ................................ ....................... 116 Option NN1: Uniform Fixed Fee Under Net Neutrality ................................ ............... 116 Option NN2: Differential Fixed Fees Under Net Neutrality ................................ ......... 117 Option NN3: Two Part Tariff Under Net Neutrality ................................ ..................... 118 No Net Neutrality ................................ ................................ ................................ ................. 120 Option NNN1: Uniform Fixed Fee Under No Net Neutrality ................................ ....... 120 Option NNN2: Differential Fixed Fees Under No Net Neutrality ................................ 122 Option NNN3: Two Part Tariff Under No Net Neutrality ................................ ............ 124 ................................ ................................ ................................ ............... 126 Three Options NN1, NN2, NN3) ................................ ................................ ............... 126 Among Three Options NNN1, NNN2, NNN3) ................................ ......................... 127 ix Options NN1, NN2, NN3, NNN1, NNN2, NNN3) ................................ ................... 128 ................................ ......................... 130 reference for Pricing Structure Under Net Neutrality (NN1, NN2, NN3) ................................ ................................ ................................ ................. 130
7 (NNN1, NNN2, NNN3) ................................ ................................ ............................. 130 NNN1, NNN2, NNN3) ................................ ................................ .............................. 131 ......................... 131 7 CONCLUSIONS ................................ ................................ ................................ .................. 135 Summary of Major Findings ................................ ................................ ................................ 135 Future Research ................................ ................................ ................................ .................... 139 APPENDIX A LIST OF NOTATIONS ................................ ................................ ................................ ........ 142 B PROOFS OF PROPOSITIONS AND OTHER ANALYTICAL RESULTS ....................... 144 Proof of ................................ ................................ ................................ ..................... 144 Proof of Proposition 3 1 (The Results of the Short Run Problem) ................................ ...... 144 Proof of a Necessary and Sufficient Condition for the Existence of a Unique ... 151 Proof of ................................ ................................ ................................ ................... 152 Proof of Proposition 3 ................................ .. 153 Proof of Proposition 3 ................................ ......... 155 Proof of Proposition 3 Optimal?) ................................ ................................ ................................ ........................... 156 Proof of Proposition 4 1 (Market Coverage Under Net Neutrality) ................................ ..... 157 Proof of Lemma 4 1 (Feasibility Condition for Outcome 2) ................................ ................ 160 Proof of Lemma 4 2 (Optimization Condition for Outcome 2) ................................ ........... 160 Proof Proposition 4 2 (Outcome 3 Dominates Outcome 2) ................................ ................. 161 Proof of Lemma 4 4 (Optimization Conditi on for Outcome 3) ................................ ........... 161 Proof of Proposition 4 3 (Outcome 4 Dominates Outcome 1) ................................ ............. 161 Proof of Proposition 5 1 (Equilibrium of the Game with a Vertically Integrated BSP) ...... 161 Proof of the Results in Table 5 1 ................................ ................................ .......................... 167 Proof of Proposition 5 2 (The Welfare Impact of Net Neutrality with a Vertically Integrated BSP) ................................ ................................ ................................ ................. 171 Proof of Proposition 5 3 (The Welfare Effect of Vertical Integration) ................................ 173 Proof of Proposition 5 ................................ ................................ ....................... 1 75 Solution of Formulation ( 6 7 ) Option NN3 ................................ ................................ ....... 181 Proof of Proposition 6 ... 186 Proof of Proposition 6 Neutrality) ................................ ................................ ................................ ......................... 187 Proof of Proposition 6 cture) ......................... 188
8 Proof of Proposition 6 Net Neutrality) ................................ ................................ ................................ .................. 192 Proof of Prop osition 6 ......... 192 Proof of Proposition 6 ........................ 193 LIST OF REFERENCES ................................ ................................ ................................ ............. 194 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 198
9 LIST OF TABLES Table page 2 1 Literature review ................................ ................................ ................................ ................ 26 3 1 Delays under No Net Neutrality ................................ ................................ ........................ 55 3 2 ................................ ................................ ................................ 56 3 3 Summary of results of the game ................................ ................................ ........................ 56 3 4 Comparison of various economic outcomes of interest under NN and NNN .................... 57 3 5 Comparative statics with respect to capacity ................................ ................................ 58 4 1 Parameter values and step sizes ................................ ................................ ......................... 81 5 1 ................................ ................................ ............ 107 5 2 Comparison of various economic outcomes of interest under NN and NNN .................. 108 5 3 ................................ ...................... 109 6 1 ................................ ................................ ...... 133
10 LIST OF FIGURES Figure page 2 1 Mapping the net neutrality debate to the underlying Internet archi tecture ........................ 25 3 1 Schematic of the model ................................ ................................ ................................ ...... 53 3 2 The content providers and their share of consumers with full market coverage ............... 53 3 3 The sequence of events in the game ................................ ................................ .................. 53 3 4 Graphical representation of the regions for arriving at different equilibria of the game when ................................ ................................ ................................ ...................... 54 3 5 General ized representation of the regions for arriving at different equilibria of the game ................................ ................................ ................................ ................................ ... 54 3 6 ................................ ................................ ................... 55 4 1 Consumers and the market shares of the two content providers ................................ ........ 81 5 1 Market structure with vertical integration of content and broadband services ................ 104 5 2 The sequence of events in the game with vertical integration ................................ ......... 104 5 3 The equilibria with vertical integration ................................ ................................ ............ 105 5 4 space integration with G or Y. ............................ 106 5 5 The effect of vertical integration on social welfare ................................ ......................... 107 6 1 ................................ ................................ ................................ ......... 132
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FOR WHOM THE NET TOLLS: A TWO SIDED MARKET ANALYSIS AND PUBLIC POLI CY IMPLICATIONS FOR THE NET NEUTRALITY DEBATE By Hong Guo August 2009 Chair: Hsing Kenneth Cheng Cochair: Subhajyoti Bandyopadhyay Major: Business Administration Net neutrality is a widely debated policy issue that could fundamentally alter the dynamics of providing and accessing online content through the Internet. As opposed to the status quo of Internet data transmission where requested packets are delivered from the local broadband service provider (BSP) to online users with no discrimination, severa l BSPs have proposed charging online content providers for priority content delivery. Opponents of the proposal proposed priority delivery and pricing mechani sm. My dissertation provides a comprehensive analysis of the net neutrality debate. I propose a technological economic framework of net neutrality and analyze two different forms of network discrimination: content provider discrimination and user discrimin ation. Using game theoretical models, I study five critical economic consequences of the potential net neutrality regulation: consumer surplus, social welfare, capacity expansion, market coverage, and vertical integration. Major findings include: If conten t provider discrimination replaced net neutrality, the monopolist BSP gains while content providers are worse off. Consumer surplus and social welfare increase or remain unchanged. In most cases, the BSP has higher incentive to expand its infrastructure ca pacity (bandwidth) under network neutrality. While content provider
12 innovations at the edge. When a BSP is vertically integrated with a content provider, social w elfare may increase or decrease depending on how effectively the vertically integrated firm generat es revenue from its content compared to the competi ng independent content provider For user discrimination, the impact of net neutrality depends on the char acteristics of the Internet data consumption market With net neutrality the BSP would prefer to charge a two part tariff for Internet access. W ithout net neutrality, a BSP may choose to charge a uniform price and degrade heavy users or charge a higher pr ice to heavy users for preferential delivery of their data packets. Interestingly, w ithout net neutrality, degrading the experience of heavy users a practice that the FCC recently banned increases social welfare. The se findings have important policy im plications and shed light for regulator s to determine the fate of net neutrality.
13 CHAPTER 1 INTRODUCTION What is Net Neutrality? Net neutrality is a wide ly d ebat ed policy issue that could fundamentally alter the dynamics of providing and accessing online content through the Internet Many of us have opinions regarding the subject of ne t neutrality, and if the public discourse on the issue is any indicator, our opinions have definite ideological overtones: do we want as the proponents of net neutrality would frame the issue the Internet to be a neutral carrier of information packets, without any regard to what the packets contain, where it originated from or where might it end and therefore want legislation to enforce that principle? Or if the question is posed by supporters on the other side of the debate do we want the free ma rket to decide what services should be charged and how, and who should be charged for these services ? The debate started and received widespread media attention when some broadband service providers (BSPs) like Verizon, Comcast and AT&T (among others) pr oposed charging popular websites for preferential access to their residential and commercial customers (Helm 2006; Waldmeir 2006) The proposal encountered stiff resistance from those who were to be charged and t hus erstwhile competitors like Google, Yahoo! and Microsoft (and seemingly disparate organizations like the Christian Coalition of America and the American Civil Liberties Union) were soon lobbying before the United States Congress to pass legislation that would prevent the broadband service providers from carrying out their proposed plan (WSJ 2006) and thereby maintain what was 1 itself is attributed to the Columbia Law School professor Tim Wu). This wo uld involve the 1 or platform.
14 2 from using their power over the transmission technology to negatively affect competition in complementary markets for (van Schewick 2 007) (Wu 2003) and while a formal definition of the operationalization of the principle does not exist, Hahn and Wallsten (2006) Internet access, do not favor one conte nt provider over another, and do not charge content Lobbying by both sides of the issue has been intense, and the United States Congress is currently considering proposals to introduce n etwork neutrality legislation (Dunbar 2006; McCullagh and Broache 2006; Windhausen 2006) The U.S. House of Rep resentatives and the Senate have held several hearings on the subject (Representatives 2005; Senate 2006) The Federal Trade Commission has also chimed in, and has recently published a report that has advised a wait and watch approach on the matter (FTC 2007) The exposition so far might lead a reader t o think that the net neutrality issue is relevant only within the United States, but as the Internet is a global entity, such a conclusion would be misleading. While the net neutrality debate initially started in the United States, the intensity of the dis cussions has led other countries across the world to consider its repercussions. Following the lead of US broadband service providers, B SPs like Deutsche Telekom in Germany and Telecom Italia in Italy are lobbying the European Commission to allow charging content 2 The terms Internet service providers (ISPs) and broadband serv ice providers (BSPs) are oftentimes used interchangeably in reality and thu s in the remainder of this dissertation
15 (EC 2006) In several other countries like Canada or Japan, the concerned regulatory agencies have begun to study the issue and its implications, since any policy directives emerging from the United States can serve as a precedent for other countries. The Two Sides of t he Debate Both sides in this debate, it would seem, have good arguments to back up their claim. Backers of the net neutrality principle wou ld like to maintain the status quo, whereby the B SPs do not look into the packets transmitted through their network a nd therefore no discrimination exists Independent commentators have asked the Federal Communications Commission to impose rules on the B SPs that would prevent them to discriminate against the third party content providers (Coalition 2002; Wu and Lessig 2003) This, they claim, would preserve the egalitarian philosophy on which the Internet was founded (Lessig 2001) Other backers of the principle are many online start ups, who claim that it would be almost impossible for them to pay these proposed fees when their revenue streams are almost non existent, since they have to give away mo st of their content for free in order to build a loyal customer base (Sydell 2006) There are venture capitalists who have argued that abandoning net neutrality would result in would be entrepreneurs becoming more hesitant to start a business, a state of affairs that might even hurt the competitiveness of the American online firms in the long run (Sydell 2006; Wu 2006a) Vint Cerf, the renowned computer scientist who is commonly referred to as one of the Internet resembling more and more like the controlled mass media today, with a few broadband service providers control ling what the customers effectively have access to (Waldmeir 2006) Tim Berners Lee, the founder of the World Wide Web, also favors keeping net neutralit y in (Berners Lee
16 2006) Finally, some people have voiced their fears of the Internet service providers starting to offer competing services ( like Internet telephony ) to their consumers at rates that undercut other rival providers, wh ich can effectively smother competition (Senate 2006) This again might result in stagnation in what has so far remained one of the most open of marketplaces. Opponents of network neutrality legislation have denied its need in the first place (National Cable & Telecommuni cations Association 2003; Owen and Rosston 2003; Yoo 2006) since they feel that broadband service providers do not have an incentive to discriminate against content providers (Speta 2000a; Speta 2000b) Broadband servic e providers themselves have argued that it is they who have put their resources to maintain and upgrade the physical services that they resources (Waldmeir 2006) deals to give (Krim 2005) With online content increasing exponentially over the years, and consumers increasingly becoming used to broadband access, the priority delivery c harges will be necessary to meet the rising costs of increasing the capacity and serving the expanded consumer base. Not having these sources of revenue might act as a disincentive to upgrade their infrastructure and affect the increasing their existing capacities. That, in turn, would affect many emerging online services like real time broadband video that by design require preferential treatment of their packets. Some B SPs contend that the new payment mechanisms might herald th e beginning of new integration of new features and services by broadband network operators is an essential part of
17 the innovation strategy companies will need to us e to compete and offer customers the services (Thierer 2004) The issue of the incentive to expand broadband capacity is de finitely of great interest to observers who have noted the gradual decline of the position of the United States in the ranks of countries that lead in broadband deployment (Yang et al. 2004) investment in telecommunicati ons as a percentage of Gross Domestic Product (0.169% as compared to the (ITU 2006) per capita broadband subscription (falling from fourth among all nations in 2001 to twelfth in late 2006 to fifteenth in April 2007) (Ricadela 2007) and average broadband speeds (the average download speed in the United States is around 4.8 Mbps, compa red to the leader Japan where it is 61 Mbps) (Cauley 2007) If the BSP s can utilize the extra profit from charging the content providers to upgrade the broadband infrastructure, it would indeed be a welcome development. It is to be noted here that in all the countries that currently lead in the va rious measures of broadband deployment none of the BSP s have implemented any mechanism for priority delivery of online content. on (Wu 2006b) and an extensive discussion of the issues might be found in (Wu 2003) In brief, network neutrality aims to address concerns raised by some specific behavior of the broadband service providers: ( 1 ) blocking of some content providers; ( 2 ) preferential treatment of one provider over another and ( 3 ) transparency failures, whereby a broadband provider fails to notify its customers and content providers what service they offer in terms of estimated bandwidth, late ncy, etc. (Wu 2006b) The current proposals by the broadband service providers has raised concerns around the second issu e i.e., the possibility that one content or application provider pays the broadband service provider for preferential
18 treatment of its packets, as the latter acts effectively as a gatekeeper between the content providers and the customers it serves. The entire debate has raised a number of unanswered questions that are of interest to researchers and practitioners alike, not to mention the regulatory agencies. The stakes involved in the issue was brought into focus during a House Committee hearing in April 2006 (Wu 2006a) I infrastructure. It has become as essential to people and to the economy as the roads, the electric grid, or the teleph The debate over net neutrality has so far been argued mainly from two angles the legal perspective, which considers access to the Internet as a b asic right that should be legislated; and the regulatory perspective, which considers any initiative to legislate pricing by a private entity as a hindrance to market mechanisms that ultimately hinder social welfare. In sharp contrast to much of the lofty rhetoric presented by t he two sides in the debate, my dissertation present s the issues from a technological economic perspective. Structure of the Dissertation The rest of the dissertation is arranged as follows. In Chapter 2, I propo se a technological eco nomic framework to provide a holistic view of the net neutrality debate. Based on the proposed framework, I further review the existing li terature and position this dissertation within the literature Chapter 3 propose s a two sided market model of digital content provision and consumption through the Internet and analyzes one form of network discrimination content provider discrimination Winn ers and losers are identified by compari ng the payoff for the BSP and content providers, the consumer surplus, and the social welfare under net neutrality versus payment mechani sm model without net neutrality. Within the same framework I also examine
19 I further study the impact of conten t application innovation on the cont In Chapter 5, I consider a special case of content provider discrimination when the broadband service provider vertically int egrates with content providers. Chapter 6 presents a different form of network discrimination user discrimination and examines six potential mechanisms of user discrimination Finally, I conclude in Chapter 7.
20 CHAPTER 2 A TECHNOLOGICAL ECONOMIC FRAMEWORK The Technological Perspective Despite the extensive discussion on this topic, t h e proper usage and context of the term (Wu 2006b) In this chapter I frame the net neutrality debate from the technological and economic perspectives. From the technological perspect ive, the five layer Internet architecture defines the underlying data transmission mechanism to end Starting from the bottom up, the physical layer defines the transmissions through the physical media. On top of this, the link layer is responsible for router to router transmissions. Following next is the network layer which establishes the logical connection between the source and the destination. Communications between the individual processes on the source and destination are defined by the transport layer. And finally, the application layer completes the data transm ission between various applications running on the end systems. The status quo of data transmission on the Internet is a simple blind mechanism which treats all data packets equally resulting in no discrimination. However, t he technology to discriminate pa ckets and streamline Internet traffic has been available at a minimal fixed cost. It is feasible to adopt a more sophisticated mechanism which has the capability to recognize and distinguish the transmitted data packets based on its source and destination, content, and/or application. As a result, some data packets may be potentially favored over others resulting in four basic forms of discrimination application discrimination, content discrimination, content provider discrimination, and user discriminati on in the context of online content provision and consumption through the Internet. Based on the required information for discrimination, there is a natural mapping from these discrimination mechanisms to the five layer In ternet architecture.
21 As shown in F igure 2 1, no discrimination (net neutrality) corresponds to the two lower layers since the physical and link layers move data from a router to its adjacent router without the information about the source and the destination of the route. Potential discrim ination (no net neutrality) corresponds to the three upper layers. Specifically, network layer contains the sender and the receiver information and thus corresponds to content provider discrimination and user discrimination The transport and application l ayers recognize processes and applications on the end systems and therefore correspond to content discrimination, application discrimination, or both. Among the four possible forms of discrimination, the current proposal by the B SPs charge popular websi tes for preferential access to their customers is content provider discrimination which is also the focus of my dissertation and I will elaborate more in Chapter s 3 4 and 5 In C hapter 6, I analyz e another form of network discrimination user discrim ination The Economic Perspective From the economic perspective, t he entire net neutrality debate has raised a number of unanswered questions that are of interest to researchers and practitioners alike, not to mention the regulatory agencies. T he regulator y agencies wou ld like to know who are the winn ers and losers if the principle of net neutrality is abolished. Specifically, if the social welfare increases as a result of abandoning net neutrality and more specifically, the end consumers are better off, the idea for the proposed payment mechanisms would gain traction among the policymakers; conversely, if abandoning the principle of net neutrality results in helping to extract more rent for only a few private agencies, the idea would find a much less symp athetic audience. Another issue of interest is to check the veracity of a key claim of the BSP s that under net neutrality, the incentive to expand the capacity of the existing infrastructure for the next generation of broadband services is much less as c ompared to when they are allowed to charge the online content providers for preferential treatment. For policymakers, this is indeed a key
22 issue. Higher capacity broadband services will enable many services that are deemed important for the society as a wh ole. Some examples of such services include disaster recovery, remote medical supervision and the like. For content providers, the next generation broadband services will enable instant delivery of high definition movies, consumer interactivity, a richer o nline shopping experience and so forth, and in the process open many new channels of revenue generation. Furthermore new start up firms may not be able to compete with incumbent firms and may be un likely to win the prioritization auction, lea ding to less innovation It is also interesting to consider the effect of the broadband service provider as a potential competitor to the content (or other service) providers. Such a situation already exists today (albeit with limited success so far) with broadband ser vice providers like Comcast building their own modest Internet portals, or with providers like AT&T or Comcast offering VoIP digital phone services (Krim 2005) T he B SPs can impose preferential treatment for their own content a nd applications over those of other providers. Policymakers would then like to ensure that the monopolist broadband service provider does not discriminate against or downgrade other content providers and enjoy unfair competitive advantage. Answers to these critical econ omic issues will help shed light for policy makers to determin e the fate of net neutrality With the background in the underlying technology and the associated economic issues, IS researchers can bring forth a holistic perspective that has of ten been lacking in the net neutrality debate. This dissertation represents one such attempt. Literature Review Formal analytical studies on net neutrality are very limited, regardless of the extensive discussions and debates on the topic. Based on the pro posed technological economic framework of the net neutrality debate, I review the existing literature in net neutrality. I summarize the
23 literature review in Table 2 1 based on types of discrimination and economic consequences examined in these studies. He rmalin and Katz view the potential net neutrality regulation as product line restrictions on the providers of last mile Internet access services (Hermalin and Katz 2007) They found that a single product restriction typically but not necessarily reduces the overall social welfare. Economides and Tag utilize a two sided market model to exam ine the effect of two sided pricing with content provider discrimination with emphasis on the cross group network externality between applications and consumers (Economides and Tag 2007) They find that net neutrality regulation increases total industry surplus in the presen ce of a monopoly BSP or in the duopoly setting. Jamison and Hauge (2007) study the impact of a premium service for content providers without net neutrality and find the provision of premium service stimulates application innovations at the edge of the Internet. Musacchio et al. (2008) propose a two sided market social welfare effect of two sided pricing and one sided pricing and find that the result is determined by the ratio between adver tising rates and end user price sensitivity In my dissertation, I examine both content provider discrimination (Chapters 3 5) and user discrimination (Chapter 6) Specifically, I propose a game theoretical model of content provider discrimination (Chapter 3) to identify the winners and losers of implementing these different types of transmission discrimination and pricing mechanism and present the ir implications on the consumers and social welfare. Within the same framework, I also investigate whether broa dband providers have lesser incentive to expand capacity (bandwidth) for the consumers if network neutrality legislation is in place. I further extend the model to study the special cases of content
24 provider discrimination: content provider discrimination with partial market coverage (Chapter 4) and vertical integration (Chapter 5).
25 Figure 2 1. Mapping the n et n eutrality d ebate to the u nderlying Internet a rchitecture ( Technological perspective of the net neutrality debate ) Technological Perspective of the Net Neutrality Debate Potential Forms of Discrimination Five Layer Internet Architecture Application Discrimination Content Discrimination Content Provider Discriminatio n User Discrimination No Discrimination Application Layer Transport Layer Network Layer Link Layer Physical Layer Net Neutrality No Net Neutrality
26 Table 2 1. Literature revie w The Net Neutrality Debate Economic Consequences Consumer Surplus Social Welfare Application Innovations Capacity Expansion Vertical Integration Different Forms of Discrimination Application Discrimination Content Discrimination Content P rovider Discrimination Economides and Tag 2007; This dissertation Economides and Tag 2007; Hermalin and Katz 2007; Musacchio et al. 2008 ; This dissertation Jamison and Hauge 2007 Jamison and Hauge 2007; Musacchio et al. 2008 ; This dissertation This dissert ation User Discrimination This dissertation This dissertation This dissertation
27 CHAPTER 3 CONTENT PROVIDER DIS CRIMINATION The status quo of prohibiting broadband service providers from charging websites for preferential access to their customers the bedrock principle of net neutrality is under fierce debate. In this chapter, I study the focal point of the net neutrality debate content provider discrimination. Specifically I develop a game theoretic model to address two critical issues of net neutrality: (1) Who are winn ers and losers of abandoning net neutrality; and (2) Will broadband service providers have greater incentive to expand their capacity without net neutrality? The Model To analyze the problem at hand, I consider a stylized model with three types of players: ( 1 ) a monopolist broadband service provider that not onl y serves consu mers in a specific geographic market by providing them with Internet access but also serves content providers by delivering their content to the consumers in this market; ( 2 ) two competing content providers that provide their service for fre e to the end users as they generate revenues from advertisers and associated ; and ( 3 ) consumers who consume content from their preferred content provider through the Internet access provided by the local BSP (as shown in Figure 3 1) I develop the model both under net neutrality (hereafter shortened to NN) and when net neutrality is abolished in favor of the regime where the B SP can charge the content providers ( i.e., no net neutrality, or NNN for short) T he advertisemen t supported revenue model I consider here is overwhelmingly prevalent among online content providers which is very different from traditional online retailers like Amazon.com. The specifics of the revenue model change from one provider to another, but in general terms, it involves no upfront fees from the customer, but rather a customer value is encapsulated in the entire lifecycle of relationship with the firm. The idea was first
28 proposed by the noted journalist and commentator on digital tec hnolo gies, Esther Dyson (1994) and is now considered mainstream by industry observer s like Chris Anderson (2008) and Nicholas Carr (2008) economics researchers like Paul Krugman (2008) and Hal Varian (McKinsey McKinseyQuarterly 2009) and I S researchers like Eric Clemons (Knowledge@Wharton 2008) In this framework, the firm gets its revenues not directly from a particular customer, but rather in a stochastic sense when these customers indirectly generate revenues through a variety of means such as banner advertisements, affiliate revenues, rental of subscription lists, sale of aggregate information, to name a few. W ithout loss of generality I norm alize the total number of end consumers (i.e. the total number of consumers served by the monopolist B SP) to 1. This unit mass of custom ers is uniformly distributed on line segment in terms of their ideal content T here are two competing online cont ent providers, Y and G where content provider Y is located at zero, while content provider G is located at the opposite end of the interval (see Figure 3 2 ). Let be the marginal consumer that is indifferent to the content between Y and G Then, the market shares for Y and G are and respectively. This amounts to the online content market of end consumers being fully covered by content providers Y and G This assumption of full market coverage will be relaxed in Chapter 4. B oth content providers offer their basic services at no cost to the end users. 1 In this model, I consider the revenue generation of the content provider as the average revenue generated (from all sources) per packet requested by the end consumer. Let and denote the revenue rate s of content provider Y and G respectively per packet for content I n other words, these two 1 In order to draw a meaningful comparison between the current and the proposed environment, one needs to choose the same revenue model for the content providers under both environments, and I have therefore chosen the revenue model that mirrors current reality.
29 parameters deno te the average rate s at which the requests for content from the consumers provide revenues to the content provider s from myriad types of advertisers that want to reach them I assume that which means that one content provider ( G ) is better than the other ( Y ) therefore can charge higher advertising fees. T his assumption does not affect the analysis and the generalized results (when and ) will be discussed later. Without net neutrality, the B SP provides preferential delivery service for content at a price which is the unit price for priority data packet transmission per packet T he technology to discriminate packets and streamline Internet traffic has been available at minimal fixed cost and therefore the cost of implementing this mechanism of priority delivery of some content is assumed to be negligible In respons e, the two content providers decide whether to pay for this preferential treatment. Then, the service decisions (whether to choose the preferential delivery service) for the two content providers can be represented by the indicator functions and Let be the Poisson arrival rate of content requested by each consumer, and it is expressed in packets per unit of time. Content provider Y and content provider G is The demand for content providers and depend on their service choices and I will provide further analyses of the demand realization later.
30 A consumer s net utility from using the services of either Y or G depends on of preferred provider ( i.e., either Y or G ) from ideal and the cost of delay that is a result of the general congestion in the access network between the BSP and the consumer. Parameter measures of the deviation from a work (Hotelling 1929) Following Mendelson (1985) I denote to be the gross value function of this content for each consumer, assumed to be twice differentiable and strictly concave. Furthermore, consumers get a congestion disutility due to waiti ng for packets : the delay cost parameter multiplied by the expected time in such a queuing system The B SP charges a fixed Internet access fee per unit time to the consumer s for Internet access Both the fixed fee and the priority charge are expressed in the same unit of time. Therefore the utility function for an arbitrary consumer is if the content provider is Y and is if the content provider is G In order to determine the delay I consider to be the capacity that the B SP provides to t he consumers, expressed in packets per unit of time. This capacity constraint affects the service that the B SP renders to th e consumers in a unique fashion. S pecifically, one can think of the packets requested by the consumers as being serviced in an M/M/1 queuing system. I assume that customers are homogeneous in terms of having the same rate of requests for content, valuation of content, and sensitivity to delay Currently under NN the congestion delay is the same for all consu mers and does not figure into the decisions. However, under NNN, the congestion delay plays an important role. To understand the impact of abolishing net neutrality, consider a situation where the BSP starts charging content providers Y and G fo r preferential treatment of their packets and suppose
31 without loss of generality that only Y decides to pay for the service. As a result, any packet from Y that is received by the broadband service provider as a request from one of its customers now gets p referred treatment to the top of the queue (these packets still face the congestion from other similarly preferred packets from Y ). I model the congestion in the network after Bandyopadhyay and Cheng (2006) and Mendelson (1985) Packets from G do not rec eive any preferential treatment and at any point in time are in fact queued after any packet from Y that might be requested at that point of time. Depending on the number of Y s in the channel, som e consumers that previously preferred G might now find the congestion of G causing enough disutility that they might now prefer Y packets are thus dependent on their service choices, denoted by Table 3 1 gives the delays of the four different outcomes under NNN for a two class priority M/M/1 queue with service preemption The in the delay expressions a re the corresponding marginal consumer who is indifferent between Y and G Individual consumers then choose the content provider that yields the higher utility. I assume that the consumer located at the two ends of the market is loyal to their correspondin g content providers as the consumer at the end point receives content of perfect fit. That is, an expression that can be simplified to This assumption ensures the existence of a meaningful competition between the two content providers G and Y one that is similar to a standard assumption in two sided markets literature, e.g., Equation (8) of Armstrong (2006) and assumption A3 of Armstrong and Wright (2007) Armstrong (2006) notes that ufficient condition for a market (p. 674).
32 I assume that the Internet service provider captures all end consu mers in under net neutrality Further, the B SPs have stated that their intention is not to degrade the online experience for any current broadband subscriber even if net neutrality is abolished and therefore we assume that the Internet service provider continues to serve all the current consumers when they start charging content providers fo r preferential delivery of their packets. In other words, we is sufficiently high that the utility for the indifferent consumer in any of the outcome s that follow is nonnegative. I consid er a monopolist B SP that delivers digital content from its local switching office to the end users. While the monopoly assumption is a simplification in some locales unlike many other countries, the extent of competition in the local broadband services ma rket is very limited in the United States, so much so that in many places, a single broadband service provider is often a de facto monopolist (Economides 2008; Hausman et al. 2001) The situation is aggravated by the high switching cost of long term service contracts and incompatible broadband technologies between cable and phone companies. Further, many customers are not qualified for a digital from phone companies, because they exceed the three miles distance limit from the phone operators the de facto monopolistic broadband service provider in several local markets (Turner 2007) Thus, in addition to providing the benefit of making the analysis tractable, the assumption closely reflects the reality of local br oadband services in the U.S. The B SP charges consumers a fixed Internet access fee per unit time. If it is allowed to charge the online content providers, the BSP would charge the content provider a price a per packet charge for the priority tran smission of its packets To keep the model tractable, I do not consider differential pricing that the BSP might employ (Shapiro and Varian 1998) Then the payoff function for the BSP is
33 under net neutrality and without net neutrality. A list of all notations is provided in Appendix A As shown in Figure 3 3, the timing of the game is as follows. The B SP ann ounces the Internet access fee to consumers and th e preferential delivery charge to co ntent providers under NNN. Based on the announced fees, the two content providers decide simultaneously whether to pay the premium price for priority delive ry of their content. After the B SP and the content providers make their respective decisions, consum ers choose either content provider Y or content provider G. In the following sections, I use backward induction to deduce the subgame perfect Nash equilibria of the game with and without net neutrality regulation. Net Neutrality In this section I analyze t he model under net neutrality where the B SP decides on the optimal Internet access f ee and consumers choose between content providers Y and G. The content providers do not have any decision to make here. Content Decisions for C onsumers Although individual consumers choose contents between Y a nd G independently, the with all the consumers located in c hoosing Y and all the consumers located in choosing G T he subscript NN denote s the case of net neutrality. The n denotes the marginal consumer that is indifferent between content provider Y and content provider G under NN and can be specified by ( 3 1 )
34 This leads to implying equal market share for the two content providers. The payoff to content provider Y is and the payoff to content provider G is Pricing D ecisions for the BSP Under NN, Interne t access fees collected from consumers are the only revenues for the BSP. Assume the B solves the profit maximization problem as follows: ( 3 2 ) This leads to No Net Neutrality Next, I analyze the situation when the B SP is all owed to charge the content providers for B SP charges the content provider should its competitor c hoose to not pay. When both content providers pay the price both their p ackets receive equal treatment. Content Decisions for C onsumers and conten and consumers decide on their preferred content provider. here are essentially four possible outcome s : neither content provider pay ( ) ; one content
35 provider pays and the other does not (which result s in two different outcome s and ); and both content providers pay ( ) Outcome 1: Both content providers opt for not paying the priority price ( ). The indifferent consumer is signaled by w hich leads to 2 Notice that this outcome amounts to the same result as in net neutrality (NN). Outcome 2: Content provider Y pays while content provider G chooses not to pay ( ffic. The marginal consumer who is indifferent between content provider Y and content provider G under NNN in Outcome 2 is specified by: ( 3 3 ) This leads to (see Appendix B for the proof 3 ) meaning that content provider Y enjoys a larger market share at the price of priority queue with preemption. 2 To facilitate understanding, I have made the numerical subscripts (1 thr ough 4) of the various variable s correspond to the differen t Outcomes 1 through 4. 3 All proofs are organized in Appendix B.
36 Outcome 3: This case is the opposite of Outcome 2. Content provider G decides to pay the preferential packet treatment pric e of per packet, while content provider Y chooses not to pay ( ). Carrying out a similar analysis, I denote the marginal consumer by who is indiffere nt between content provider Y and content provider G under NNN in Outcome 3 and is specified by: ( 3 4 ) It follows that meaning that content provider G enjoys a larger market share at the price of Outcome 4: Both content providers pay the priorit y pric e to have their content delivered ( ) Delivery S ervice D ecisions for C ontent P roviders Given ce rtain values of and content providers Y and G decide whether to pay for by represented by Table 3 2 presents the payoff matrix for the content providers under the four outcomes.
37 Outcome 1: The revenues for content provider Y and G are and respectively. The incentive compatibility constraint for content provider Y is and the incentive compatibility constraint for content provider G is Outcome 2: Th e revenue for Y is and the revenue for G is The incentive compatibility constraint for content provider Y is and the incentive compatibility constraint for content provider G is These two incentive compatibilit y constraints can be reduced to ( 3 5 ) and ( 3 6 ) C ompare the right h and side (RHS ) of constraints ( 3 5 ) and ( 3 6 ) leads to since and Thus, there is no feasible such that content provider Y pays for the preferential delivery while content provider G does not. Therefore Outcome 2 cannot be an equilibrium. Note that this result is driven by the fact (or more correctly, the assumption) that In other words, if we can never have a n outcome where content provider Y decides to pay and G does not. The assumption does n ot affect the key results of the analyses. If however, this assumption is reversed, then Outcome 3 instead of Outcome 2 cannot be an equilibrium
38 Outcome 3: The revenues for Y and G are and respectively. The incentive compatibility constraint for content provider Y is and the incentive compatibility constraint for content provider G is These two incentive compatibility con straints can be reduced to ( 3 7 ) and ( 3 8 ) Comparing the RHS of ( 3 7 ) and ( 3 8 ) gives To analyze the magnitude of this ratio, we consider two possibilities in the relative values of and : Case I: ; and Case II: If Case I holds there is no feasible such that content provider G pays for the preferential delivery and content p rovider Y does not, and therefore Outcome 3 cannot be an equilibrium If Case II holds Outcome 3 may be an equilibrium I will further analyze the equilibrium results in next subsection (i.e., stage 1 of the game). Outcome 4: In this case, content provide rs Y and G get the same revenues ( i.e., and for Y and G respectively) as those in Outcome 1. Both content providers, however, incur an extra expense of which paid to the BSP The incentive compatibility constraint for content provider Y is and the incentive compatibility constraint for content
39 provider G is been described in other contexts (Brander and Spencer 1983; Roller and Tombak 1990) both content providers know that they would be better off by not paying, but given the relative proximity of their per consumer revenue s tre ams, they end up paying the B SP. Pricing D ecisions for the BSP Expecting the best response s of both content providers and consumers the B SP analyzes the maximum profit that he can make under the various permutations of the choices of the content providers to pay him (recall that the content providers can decide either to pay or not pay marginal consumer Outcome 1: Both content providers o pt to not pay the priority price ( ). Similar to the NN case, the Internet access fees collected from consumers is the only revenue for the B SP. The BSP solves the following profit maximization problem: ( 3 9 ) The first two constraints are participation constraint s for consu mer s of content provider Y and G respectively The last two constraints are incentive compatibility constraints for the and Since (this is shown to be true in Outcome 3, which is
40 discussed later) and which implies that the B preferential delivery charge can be specified as Therefore the results of the prof it maximization problem of the B SP in Outcome 1 are: the B SP makes a profit of and charge s the content providers a fee such that The best response for both content provider Y and content provider G is to Not Pay the fee Outcome 2: Content provider Y pays while content provider G chooses not to pay ( ) In addition to the revenue from end users the B SP also gains revenue from content provider Y for preferential delivery of Y The b roadband p profit maximization problem is thus: ( 3 10 ) The first two constraints are the participation constraints of the consumers that prefer G and the consumers that p refer Y respectively. The last two constraints ensure that while content provider Y will pay, content provider G will not. As discussed in the second stage of the game, there is no feasible to induce Outcome 2. Outcome 3: Content provider G pays while content provider Y chooses not to pay ( ) T he BSP :
41 ( 3 11 ) ensure that this outcome actua lly holds i.e., Y does not pay, but G does. Recall the two cases If Case I holds i.e., there is no feasible and Outcome 3 is not possible. If Case II hold s, i.e., the BSP optimal choice of pricing strategy and its profits are given by the expressions and Outcome 4: Both content providers pay the priority pric e ( ), so that neither content gets any relative advantage for delivery. The B SP now gains revenue from consumers as well as both content providers and therefore solves the following optimiz ation problem: ( 3 12 ) It follows that the BSP is given by and while his profit is
42 We note that ( i.e., Case II) is a necessary condition for Outcome 3 to be an equilibrium When this condition is not satisfied, i.e., (Case I), we have only two potential equilibria ( Outcomes 1 and 4) Recall further that Outcome 2 is never an equilibrium as long as In order to determine his optimal pricing st rategy the BSP compar es the profits under the various outcomes and for a given set of parameter values, choose s its pricing strategy (which drives the equilibrium to one of the above mentioned outcomes) to arrive at the highest profit As the monopolist gatekeeper between the content providers and the customers, the BSP possible profits. We can readily observe that in both Case I and Case II. As a result, in Case I, the broadband provider will set the final to and realize the profit since Outcomes 1 and 4 are the only two potential equilibria In Case II, t he B SP needs to compare with in order to determine the outcome that gives the maximum profit, which leads to the following comparison : after applying Equation ( 3 4 ) and some algebra.
43 We observe that if then The broadband provider will then set the menu of prices to and will realize the profit Conversely, if then The broadband provider will then set the menu of prices to and to attain a profit of Note that the condition that ensures the B S P to realize more profit in Outcome 3 than Outcome 4 also satisfies the condition for Outcome 3 to be feasible since We thus simplify the combination of the profit comparison condition (between Outcomes 3 and 4) and the feasibil ity condition of Outcome 3 into the following two cases (as shown in Table 3 3) Case A: Outcome 3 is not only feasible but also more profitable to the B SP than Outcome 4. Therefore, Outcome 3 is the equilibrium. The required condition for Case A is ( 3 13 ) Case B: Either Outcome 3 is not feasible, or when Outcome 3 is feasible Outcome 4 is more profitable. Therefore, Outcome 4 is the equilibrium. The required condition for Case B is ( 3 14 )
44 Figure 3 4 summarizes these results graphically, by plotting against The BSP dictated by the relative magnitudes of their revenue generation capabilities. In Region A (corresponding to Case A) where the BSP chooses its pricing strategy in such a way as to drive the game to Outcome 3, where the content provider with higher profitability has the incentive to pa y for priority delivery, while the content provider with lower profitability does not have the incentive to pay that fee. C onversely, in Reg ion B (corresponding to Case B ) where Outcome 4, when both content providers pay the B SP the priority delivery fee T he choice of making is sim ply a matter of convenience of exposition, that the generalized results are symmetric on either side of the line as indicated in Figure 3 5 Regions C and D can be interpreted analogously as Regions A and B. Winn ers and Losers Comparison B etween NN and NNN In this section, I consider the resulting surpluses for the various players. These results can then be used by the policymaker who has to decide whether to allow the BSP to charge for pref erential service (i.e. opt for NNN) or continue to maintain the NN status quo The policymaker can proceed to compare the equilibrium under NN and NNN by evaluat ing the payoff for the BSP and content providers, consumer surplus, and social welfare under each regime F rom the BSP NNN is preferred to NN since the profits in either Case A or Case B or is higher than What is of interest to the policymaker is wh ether the other participants gain from this arrangement too and whether social welfare as a whole increases. T he res ults of this analysis are summarized in Proposition 3 1.
45 Proposition 3 1 : (Winn ers and losers in the short run) The economic outcomes in the short run under NN and NNN vary. Specifically, using NN as the benchmark, Social welfare would either increase or remain unchanged depending on parameter values as stated in conditions ( 3 13 ) ( 3 14 ) Likewise, consumer surplus would increase or remain unchanged. Content providers are usually worse off under NNN except under C ondition ( 3 13 ) when the content provider paying the priority delivery fee has the same surplus as under NN. The BSP is unambiguously better off. Proof: See Appendix B T hese results are o rganized in Table 3 4 Appendix B contains the details of their derivation s Clearly the gains of abolishing net neutrality are not experienced equally. While the monopolist broadband service provider gains if no net neutrality were in place (in both Cases A and B of Table 3 4 ), the content providers are definitely worse off under this arrangement. Only content provider G rplus is unchanged under Case A It is interesting to no te that G does not get to enjoy the increase in the number of consumers, since the extra rent is fully extracted by the Internet service provider. It is therefore no wonder why the content providers and the Internet service providers have been on the oppos ite sides of the net neutrality debate. The fate of the end consumers is more nuanced. If the two content providers do not differ significant ly (regions B or D in Figure 3 5 ) in terms of their revenue generation rate s the consumer surplus is unchanged. C onsumers as a whole do stand to gain if one content provider is significantly better than the other in revenue generation (regions A and C in Figure 3 5 ). This increase in overall c onsumer surplus, however, is derived at the expense of the group of consumers whose content provider does not pay the priority charge, a result contrary to the
46 assertion of the B SPs that no consumer would be left worse off under the new arrangement (WSJ 2006) S ocial w elfare as a whole, in contrast, is at least as high under NNN as it is under NN, and is sometimes higher Under NNN Case B, social welfare (like consumer surplus) does not change, but there is a transfer of wealth from the content providers to the Internet service provider. This transfer is made possible by the priority delivery charge that the Internet service provider extracts from both content providers, since the subscription fee to end users do es not change ( ). Under NNN Case A, the consumer surplus increases due for the most part to the lower subscription fees for all consumers ( ). The winners under this arrangement are the consumers of content provider G (who are a majority) and the B SP, while the lo sers are the consumers of content provider Y and content provider Y itself. Content provider G from the additional consumers that have migrated from Y is fully siphoned away by the B SP. Capacity Expansion Decision Does NN Hinder the Broadband S Incentive to Expand Infrastructure Capacity? T he other key question for the policymaker is the broadband to expand capacity under NN. To discuss this question, I consider t he long run problem where the BSP ca n choose its capacity Let be the cost associated with capacity The long run problem can be modeled as a three stage game where the BSP chooses capacity and announces the Internet access fee to consumers and preferential delivery fee to content providers under NNN in the first stage. Based on the announced fees, in the second stage content pro viders choose whether to pay or not pay for preferential delivery, and in the third stage consumers choose between content provider Y and content provider G In the long run problem, we also need to consider the cost of
47 capacity expansion (this was not an issue in the preceding analysis, since in the short run the existing network capacity is fixed). The objective is to determine the BSP optimal capacity decision under NN and NNN, and compare the two meaningfully in o rder to find the regime under which the incentive for the BSP to expand capacity is higher. Finally, the choice of regime (NN or NNN) is not under the control of the BSP and is a choice that lies with the policymakers (who can calculate the aforementioned incentives). I go through a process similar to that employed in analyzing the short run problem to investigate the BSP Decisions for B oth C ontent P roviders and C onsumers in the C apacity E xpan sion P roblem Given certain and the analysi s of the best responses for content provider s and consumers do not differ from those in the short term problem and is therefore no t repeated here run problem will be multiplied by where run profi ts are for Y and for G where represents the four different outcomes. The long run utility of an arbitrary consumer is if content provider Y is chosen and is if content provider G is chosen. Pricing and C apaci ty Expansion D ecisions for the B SP in the C apacity E xpansion P roblem Expecting the best responses of content providers and consumers I evalu ate the optimal decision for the BSP
48 Under Net Neutrality or Outcome 1: The ( 3 15 ) ( 3 15 ) where and are as defined in Outcomes 2 and 3. T he long run objective function represents the net cash flow for the BSP We know that the first constraint is binding : i.e. The optimal capacity can be derived by maximizing the long term net cash flow Equation ( 3 16 ) gives t he first order condition of this optimization problem. ( 3 16 ) The term is the marginal inc rease in profit of the BSP with respect to and as such Outcome 2: It can be proved analogously as discussed earlier in the short run problem that given the na ture of the parameter values, this outcome is not possible. Outcome 3: determined by As before, and
49 The B SP will choose the optimal capacity to maximize the long term net cash flows once again: Then can be characterized by the first order condition in Equation ( 3 17 ) : ( 3 17 ) Outcome 4: and Similarly, the corresponding net cash flows are given by and is the solution of Equation ( 3 18 ) ( 3 18 ) Comparing the B SP or ) to that under NN, the BSP has more incentive to expand capacity under NN when or ( 3 19 ) and the B SP has more incentive to expand capacity under NNN when ( 3 20 ) These conditions are graphed in Figure 3 6 with Condition ( 3 19 ) corresponding to the shaded
50 area and Condition ( 3 20 ) corresponding to th e unshaded area. P roposition 3 2 su mmarizes the resu lt concerning B Proposition 3 2 : ( The B expand infrastructure capacity) Except for the re gion defined in C ondition ( 3 20 ) the B SP has more incentive to expand capacity under net neutrality. Proof: See Appendix B To understand the nature of the result in Figure 3 6 note that the capacity costs under NN and under NNN are identical given any same capacity leve l. Hence, we only need to focus on the revenue of the BSP Further, under NN, the BSP is derived solely from the consumers while the BSP contribution from the consumers and th e contribution from the content providers. The revenue contribution from consumers increases in (as consumers enjoy reduced congestion), while the revenue contribution from the content providers decrease in (as the content providers have a reduced willingness to pay for priority delivery when congestion is reduced). Specifically, when both content providers pay (Regions B and D in Figure 3 6 ), the BSP run revenue from consumers is which increases in T he BSP run revenue from the content providers is which decreases in Thus, by increasing the capacity t he BSP BSP
51 c apacity, the BSP will almost always have a higher incentive to increase capacity under net ne utrality. When only one content provider pays (Region A or C in Figure 3 6 ), and, the contribution from the consumers is which increases in and the contribution from the content providers is which decreases in The B SP gains pacity a result similar to the case where both content providers pay. It is only in the small unshaded area in Figure 3 6 that under Condition ( 3 20 ) the gain outweighs the loss to give the BSP more incentive to expand the capacity under NNN. BSP optimal capacity choices under NN and NNN, and whether the BSP al capacity choices under NN and NNN are socially optimal. P ropositions 3 3 and 3 4 provide useful guidance to the policymaker in addressing these questions. Proposition 3 3 : ( The BSP Ex cept for the region defined in C ond ition ( 3 20 ) the optimal capacity choice under net neutrality (NN) is higher than under no net neutrality (NNN). Proof: See Appendix B. Proposition 3 4 : ( Whether the B capacity choice is socially optimal? ) The BSP always invests at the socially optimal level under net neutrality. Abolishing net neutrality results in underinvestment in infrastructure capacity by the B SP when both content providers pay the priority delive ry charge, and either underinvestment or overinvestment when only one content provider pays the priority charge.
52 Proof: See Appendix B. A corollary of Propositions 3 3 and 3 4 is that while the B might be higher under NNN than u nder NN in some specific instances, this higher capacity choice reduces social welfare T he comparative statics for the various pricing variables and the surpluses of the various parties involved with respect to the capacity are summarized in Table 3 5
53 Figure 3 1 Schematic of the model Figure 3 2 C ontent providers and their share of consumers with full market coverage Figure 3 3 The sequence of events in the game Content Provider Y Content Provider G Internet Backbone End Consumers Broadband Service Provider at local loop 0 ( Y ) 1 ( G ) marginal consumer T he BSP announces and Content providers choose Pay or Not Pay Stage 1 Stage 2 Co nsumers choos e content provider Y or G Stage 3
54 Figure 3 4 Graphical representation of the re gions for arriving at different equilibria of the game when Figure 3 5 Generalized representation of the regions for arriving at different equilibria of the game Region A Region B Region D Region A Region B Region C
55 Figure 3 6 BSP s incentive to expand capacity In the sha ded region, the BSP has higher incentive to expand capacity and the optimal capacity level is higher under NN. In the white region, the BSP has lower incentive to expand capacity and the optimal capacity level is lower under NN. Table 3 1 Delays u nder No Net Neutrality G pays G does not pay Y pays Y does not pay Region D Region A Region B Region C
56 Table 3 2 ayoffs G does not pay G pays Y does not pay Y pays Table 3 3. Summary of results of the game Case A: Case B:
57 Table 3 4 Comparison of various economic outcomes of interest under NN and NNN Note: The text in parenthesis shows how those specific economic outcomes change when moving from NN to NNN. NN (Benchmark) NNN (Case A: Only G pays) NNN (Case B: Both Y and G pay) ( Lower ) ( Unchanged ) N/A BSP Revenue ( Better off ) ( Better off ) Content Provider ( W orse off ) ( Worse off ) Content Provider ( Unchanged ) ( Worse off ) Consumer Surplus ( Better off ) ( Unchanged ) Social Welfare ( Increased ) ( Unchanged )
58 Table 3 5 Comparative statics with respect to capacity NN (Benchmark) NNN (Case A: Only G pays) NNN (Case B: Both Y and G pay) N/A BSP Revenue Content Provider Content Provider Consumer Surplus Social Welfare Legend: : increasing in ; : decreasing in ; : independent of ; : depends on parameter values.
59 CHAPTER 4 CONTENT PROVIDER DIS CRIMINATION AND MARK ET COVERAGE In all extant literature on net neutrality the focus has been either on the competition between content providers when the BSP institutes a fee for preferential delivery by assuming full market coverage ( Chapter 3 in this dissertation) or on market coverage under the assumption of no competition between the content providers (Economides and Tag 2007; Hermalin and Katz 2007) The reason for these simplifications is analytic al tractabil ity: a two sided market framework that models the externalities of limited bandwidth 1 along with intra group and inter group strategic effects between agents on both sides of the two sided platform introduces significant analytical challenges by itself. Fo r example, a standard assumption made in existing literature on two sided markets is that of full market coverage in order to ensure analytical closure (Armstrong 2006) In this chapter I relax the requirement for full market coverage and analyze the effect of the proposed net neutrality legislation on broadband market coverage. The resulting analysis thus models the three fund amental aspects of the net neutrality debate: (1) the negative externalities associated with the constrained bandwidth between the BSP and the consumer; (2) the competition between content providers that makes it possible for the BSP to charge for preferen tial delivery; 2 and (3) the possibility that the content providers might subsidiz e broadband access for the consumers, which would result in enhanced broadband market coverage. 1 A between the BSP and the consumer) that is shared by packets of all t he content providers, and the very existence of this constrained resource has negative externalities associated with it (to see the veracity of this claim, consider this BSP s would not have any credible mechanism to charge for preferential access) 2 Modeling the competition between the content providers is essential if there were no competition between the content providers, the BSPs could not have credibly instituted a charge for preferential deliver y.
60 More importantly, addressing this modeling challenge enable s us to answer three key research questions that are of interest to academics, practitioners and policymakers alike. First, how does net neutrality affect the BSP S econd, how does the BSP affect consumer surplus and social welfare ? And finally, ca n the abandonment of the net neutrality principle result in some providers being shut out of the market, thereby affecting neutrality proponents who have argued t hat in the absence of net neutrality regulation, online competition and innovation will be reduced in the long run. Answers to all these questions have significant policy implications on the future of broadband access. One novel aspect of this chapter is a methodological contribution (detailed in the section) whereby I take advantage of the structural properties of the analytical model to suggest an innovative algorithm to tackle the model which is otherwise analytically intractabl e. The Model Following the model proposed in Chapter 3, I consider a unit mass of consumers. As shown in Figure 4 1 consumers are uniformly located on with two content providers at the two ends of the line segment. For any the consumer gets utility by subscribing to content provider Y utility by subscribing to content provider G and utility by staying out of the market (i .e., by subscribing to neither). The consumer located at then compares the three options with their corresponding utilities of and and chooses the one with the highest utility.
61 As shown in Figure 4 1 and are marginal consumers for Y and G respectively, with Then the market shares for content providers Y and G are and respectively. A special case of corresponds to the case of full market coverage in which marginal c onsumers are indifferent between content provider Y and content provider G When the marginal consumers of each of the content providers are indifferent between subscribing to the preferred content provider and not subscribing t o any Internet access service at all. The timing of the game and other model setups are the same as the model presented in Chapter 3. Net Neutrality With net neutrality, the only revenue source for the BSP is the Internet access charge from consumers. Spec ifically, the BSP solves the following optimization problem: ( 4 1 ) C onst raints and in F ormulation ( 4 1 ) are the participation constraints for the consumers and constraints and are the incentive compatibility constraints for the consumers. lem yields Proposition 4 1
62 Proposition 4 1 : (Market c overage under n et n eutrality) Under ne t neutrality (1) when where i.e., the BSP covers the whole market; and (2) when there exists a unique when the BSP covers only part of the market. Proof : See Appendix B No Net Neutrality Under NNN, four possible outcomes can take place: (1) The BSP elects not to charge either content provider for preferential delivery, which is the same outcome as would occur under a net neutrality regime (we call this Outcome 1); (2) the BSP decides to charge the content providers for preferential delivery of their packets, but only content provider Y decides to pay that fee (Outcome 2); (3) analogous to the previous outcome, but now only content provider G decide to pay the preferential delivery fee. For each of these outcomes where we den ote the fixed price that the BSP charges the consumers to be and a per packet priority delivery fee (where ) to the content providers, and as a result gets a profit of The marginal consumer for content provider Y is denoted by while the marginal consumer for content provider G is denoted by Each of the se outcomes can arise in equilibrium depend ing on the magnitudes of the exogenous parameters. The monopolist BSP can calculate its profit in each of the outcomes, and depending on the magnitudes of the exogenous parameters, any one of the outcomes can conceivably produce the highest profit for the BSP. In other words, the BSP, in its capacity as a
63 two sided market platform, can effectively use its menu of prices equilibrium of the game to whichever outcome that would secure it the highest profit. We next ana lyze the different outcomes, noting that Outcome 1 (i.e. the outcome under the net neutrality regime) has already been analyzed in the previous section. Outcome 2: ( Y pays and G does not pay) ( 4 2 ) In F ormulation ( 4 2 ) c onstraints and constraints. Only when Y take effect. Similarly, only when G take effect. Notice when or constraint or would be automatically satisfied this type of a for mulation is necessary in order to ensure that when Y or G have no market share, the BSP would not have to take into account their respective participation constraints. C onstraints and represent the co Y always prefer Y over G while the
64 consumers of G prefer it over Y compatibility constraints: that is, the content provider Y prefers this out come (where it has a market share of by paying the priority delivery fee to the BSP) over Outcome 1 (and it compares its payoff in Outcome 2 to that of Outcome 1 since in that outcome G still does not p ay), where it does not pay a fee and has a market share of ; and analogously, content provider G prefers this outcome to Outcome 4 (when it does pay the BSP, and so does Y ). Note that as a result of paying the priority delivery fe e, packets from Y face an average delay only to the extent of the presence of other packets from Y while the packets of G class priority queue with preemption to depict the waiting time for th e consumers of these two content providers. For any consumer situated at data packets from Y are transmitted with higher priority with a waiting time of while data packets from G are transmitted with l ower priority with a waiting time of Lemma 4 1: (Feasibility Condition for Outcome 2) : A necessary and sufficient condition for Outcome 2 to be feasible is Proof: See Appendix B. Lemma 4 2 (Optimizati on Condition for Outcome 2): When Outcome 2 is feasible, P roof : See Appendix B
65 Outcome 3: ( G pays and Y does not pay) ( 4 3 ) In F ormulation ( 4 3 ) constraints and constraints, with the LHS of the two inequalities reflecting the fact that the BSP needs to hare is positive, i.e., only when Y take effect; and similarly, only when G take effect. Notice when or constraint or would be automatically satisfied. The constraints and represent the incentive compatibili ty constraints of the content providers (the relevant alternative outcome for Y is now Outcome 4, while that of G is Outcome 1); and and represent the fact that the consumers of each of the content pr oviders prefer it over its rival. Once again, the utility functions show the delay costs associated with a two class priority queue.
66 Proposition 4 2 : (Outcome 3 dominates Outcome 2) Outcome 3 always yields at least as high a profit for the BSP as Outcome 2 Proof : See Appendix B. Lemma 4 3 (Feasibility Condition for Outcome 3): A necessary and sufficient condition for Outcome 3 to be feasible is Proof : Similar to the proof of Lemma 4 1. Lemma 4 4 (Optimization Condition for Out come 3): When Outcome 3 is feasible, Proof : See Appendix B. Outcome 4: (Both content providers decide to pay the BSP ) We note that under this outcome, neither content provider gets a differential advantage in terms of priority delivery of its packets: in other words, the only concession they extract by paying the fee is that its packets will not get de logical question that arises is why would either content provider then decide to pay? The answer would ideally have liked to pay the BSP (and would indeed be better off if both decided not to pay), but would be in a worse off situa tion compared to the outcome where it decided not to pay but the other provider did.
67 ( 4 4 ) In Formulation ( 4 4 ) constraints and constraints. The constraints and represent the incentive compatibility constraints of the content providers (the relevant alternative outcome for Y is no w Outcome 3, while that of G is Outcome 2). Binary variables (or ) represents whether Outcome 2 (or Outcome 3 ) is feasible. These two variables are defined as: and and ensure that the constraints and come into play only when the corresponding relevant outcome is feasible. Constraints and rep resent the fact that the consumers of each of the content providers prefer it over its rival.
68 Proposition 4 3 : (Outcome 4 dominates Outcome 1) Outcome 4 as under Outcome 1. Proof : See Appendix B. C omputational An alyses Solution Procedure Based on the analytical model, we observe that Outcome 1 is independent of other outcomes However, Outcome 2 depends on (in Outcome 1) and (in Outcome 4). Similarly, Outcome 3 depends on and and Outcome 4 depends on and Due to the interleaving structure of the four outcomes, it is impossible to derive analytical results from the model even with the help of technical computing tools such as Mathematica. However, there are several unique structures of the problem that allow us to analyze it and understand the nature of the equilibria First, as stated earlier, Outcome 1 is independent. Second, Lemma 4 1 and 4 3 show that we can decompose the solution space into four components based on the feasibility conditions of Outcome 2 and Outcome 3. Third, Lemma 4 2 and 4 4 show that if we assume the feasibility of either Outcome 2 o r Outcome 3, then their respective optimal solutions would depend on ly on Outcome 1. These insights lead us to develop Algorithm 1 that allows us to solve the problem as indicated below. Note that w ithout employing th is computational solution procedure, it would be impossible to gain the various important managerial and regulatory insights into this overly complicated and considerably generalized problem. I believe that one of the main contributions of th is chapter is this particular solution methodology. A two sided platform, by its very nature, has different pricing options on the two sides it caters to. When the agents on the two sides of the platform can strategically interact among themselves in a manner that can affect their respective payoffs, the pro fit optimization
69 strategy for the platform with a particular pricing strategy (call it Option 1) will have to take into account other pricing strategies (say Option 2 or 3 and so on) that can plausibly be preferred by the agents: in other words, every poss ible outcome is intertwined with other outcomes. Such possibilities present unique challenges for the modelers, which is why they often resort to certain simplifying assumptions. I demonstrate that analyzing the structural properties of the model can often yield insights on the possible equilibria of the game, even if it is not possible to get closed and effectively solves the problem. Algorithm 4 1: Step 1: Since Outcome 1 is independent, we solve Outcome 1 first and get and Step 2: Although Outcome 2 is related to Outcome 4, the va lue of only determines the feasibility of Outcome 2. The optimal solution of Outcome 2 depends only on So we initially assume Outcome 2 to be feasible and solve for Outcome 2 based on the result of O utcome 1. Step 3: Using similar reason ing we then assume Outcome 3 to be feasible and solve for Outcome 3 based on the result of Outcome 1. Step 4: Solve Outcome 4 in four different cases: Case 1: (Both Outcome 2 and Outcome 3 are feasible) Solve Outcome 4 with all the constraints Add in the feasibility constraints for Outcome 2 and 3. Case 2: (Only Outcome 2 is feasible) Solve Outcome 4 in absence of the constraint involving Outcome 2 Add in the feasibility constraint for Outcome 2 Add in the infeasibilit y constraint for Outcome 3
70 Case 3: (Only Outcome 3 is feasible) Solve Outcome 4 in absence of the constraint involving Outcome 3 Add in the feasibility constraint for Outcome 3 Add in the infeasibility constraint for Outcome 2 Case 4: (Neither Outcome 2 no r Outcome 3 is feasible) Solve Outcome 4 in absence of the constraint involving Outcome 2 and 3 Add in the infeasibility constraints for Outcome 2 and 3 Step 5: Comparing all feasible where denotes the above four cases, we can find the optimal and Step 6: Compare to and determine the equili brium (since from Proposition 4 2) We note that even if Outcome 1 is dominated by all the other outcomes, we still need to solve for it, since the equilibrium output is used to find out the equilibria in outcomes 2 (Step 2) and 3 (Step 3). Similarly, even if Outcome 2 is dominated by outcomes 3 and 4, we still need to solve for its equilibrium in Step 2, since those results are used in Step 4 To computationally explore the different outcomes and when each of them would dominate we will need to effectively change the various exogenous parameter values so that we cover the entire parameter space. Note that while we can explore some of the parameters (specifically and as I will explain later) exhaustively for all possible values due to the nature of their impact on the solution, the other parameters can theoretically assume an infinite range of values. In the absence of a closed form solution, what is necessary therefore is that for the latter set of parameters, we cover the entire range of values that can judiciously be considered within the limits of empirical reality. That proc ess is explained in the next sub section.
71 For each of the combinations of the parameter values that I end up selecting, I then find the profit and in the manner outlined in the algorithm. This needs us to first select a param eter values, and then perturb the different parameters around these baseline values, so that they cover all possible values that can be expected to be encountered in reality. Since every iteration with the parameter values involve solving several non linea r optimization problems, we would also hav for incrementing the different parameter values, which ensures that while we effectively explore the parameter space, we do not end up having to perform an unmanageable number of optimization problems. As we detail in the discussion that follows, we ended up carrying out the solution procedure outlined in the algorithm for 10,890 sets of parameter values. The Baseline Model There are seven parameters in the analytical model. Wit h so many parameters to be baseline values and then perturb those values sufficiently to explore the different possible scenarios and outcomes. In the baseline model I set , and the rationale for which I explain next. Note that not all the parameters in the model need to be changed independent of the other parameters. For example, with respect to the parameters and what is important is not their absolute values but the u tilization rate of the service queue and hence I set the b aseline utilization ratio at 0.5 We then change the utilization ratio (by changing while keeping fixed at 1) from a low of 0.02 to a high of 0.98, increasing in step sizes of 0.12. Consider next the two parameters that estimate the revenue generation rates and Once again, what concerns
72 us is the relative magnitude of these parameters rather than their absolute values 3 To estimate the baseline values of the revenue generation rates and I decided to look at the two companies that have come to symbolize the ad sup ported revenue model in the online world that I model in this dissertation : Google and Yahoo! From the 10 Q documents of Google and Yahoo I obtained their 9 month revenue, and then convert ed these amounts into their per day revenue. From Nielsen Online, I obtained their 1 month unique audience and analogously convert ed these numbers to a per day unique audience figure for both the firms Dividing the former per day revenue amount by the per day unique audience, I was able to designate the baselines values of and Thus, when we perturb these values, we need to only change one of them I chose that to be to find the effect of differing revenue generation rates. I then conside r values of to be as low as 0.1 and as high as 3.7 ( i.e., similar to going beyond 3.9 is unnecessary since in order to capture the dynamics of the game, one only need to consider relative magnitudes of the revenue generation rates). The parameters that remain to be estimated are and and as stated earlier, they can theoretically vary within an infinite range. To find a practical range of values for these of these parameters can be kept fixed relative to the others, and here, I thought the unit fit cost would be the best choice, since one would be more interested in seeing how the different outcomes are affected by changing the other two. 3 charge both content providers of just content provider G Thus, only the relative magnitudes of and figure into
7 3 With set at the baseline value of 1, set the utilization ratio initially at 0.5. Assume now that Y has half the market share, i.e., Then, the denominator of the expression for delay is 0.5 (for a multiplica tive effect of 2). Setting and (as we have under the baseline scenario) then makes the disutility of the delay in the baseline scenario to be twice as perni cious as the disutility of the fit cost whi ch we thought to be a reasonable baseline outcome (if one considers the two content providers to be Google and Yahoo!, such an assumption states that a consumer feels relatively more aggrieved with a delay in accessing their content than with the fact that I felt that setting made the gross valuation sufficiently larger than the delay for a baseline scenario, so that many consumers still have a large enough utility in acce ssing the services of the content providers. I then set a lower bound for to be 3 (at which point the net utility would be negligible or even negative), and the upper bound at 13 (at which point it dominates any other disutility ), with a step size of 1. Keeping throughout, I then modified in effect exploring the relative magnitude of the fit cost to the delay cost. If we keep all other parameter values at the baseline figur es, keeping at a lower bound set at 0.5 gives the consumer a net utility (before accounting for access charges) of 3. That value goes down to 1 at the upper bound of at 1.5. The step size for the chan ge in these values was kept at 0.1. Thus, the total number of exploration points for the entire parameter space was 10,890. The range of the parameter values and step sizes employed for each are sum marized in Table 4 1 As menti oned earlier, each run of Algorithm 4 1 involves solving several non linear optimization problems. Since I had to run the algorithm for the 10,890 sets of parameter values, I
74 deployed the algorithm on subsets of the parameter space on three different deskt op machines running Microsoft Excel 2007 with Premium Solver Version 8 on the Windows XP operating system (64 bit and 32 bit). The configurations of the three computers were: Intel Core 2 Quad processors with 8GB RAM; Intel Core 2 Duo processor with 4GB RA M; and Intel Core 2 Duo processor with 2GB RAM. The entire series of computations took a little more than a week to complete. Research Findings The computational analysis produces several observations. These observations are important, since given the numb er of parameters in play, it is otherwise impossible to isolate the potential effect of each of them in the resultant equilibria. Even in those cases where I have managed to successfully explore the analytical solutions, the wide variety of parameters did not allow me to gauge the practical implications of the results. For example, I have proved analytically that the BSP will always prefer an NNN outcome (i.e. either Outcome 3 or 4) to the NN outcome (Outcome 1). However, this finding does not imply that the BSP would always opt for the NNN outcome in practice: for example, if the advantage gained from moving to the NNN regime is relatively small, and there are substantial administrative costs in implementing a pricing plan for the content providers (in te rms of, say, new networking equipment like smart routers that identify the origin of the packets and then correctly place them in the two tier queue without introducing any additional delay), then the BSP would probably not opt for the NNN outcomes. Furthe r, it would be interesting to explore how the other constituents, i.e. the consumers and the content providers, fare under NNN insights which we could not get using purely analytical techniques, but that are now available as a result of the computationa l analyses. For example, how would the fixed access fee compare under the two regimes? I discuss all su ch findings in the following sections.
75 The F ate of the C onsumers U nder NNN Fixed access fee : I find that out of the total 10890 scenarios, in 6287 (58%) scenarios and in the remaining 4603 (42%) scenarios. In other words, the optimal fixed fee that the BSP would charge under NNN is never higher (and often lower) than the optimal fixed fee it would charg e under NN. Further, when is smaller, the statistics of the computational analyses results indicate that it is on average nearly 21% lower than the corresponding Profit contribution from consumers: Whi le the BSP gets a percentage of its revenue (profits) from the content providers under NNN, it would be interesting to find out whether the BSP gets less revenue as a whole from the consumers under NNN than under NN. The results of the computational analys es clearly indicate that that is true: in other words, for all scenarios, and therefore the content providers I profits from the consumers under NNN (under NN, it is always 100%). Within the ambit of the results of the computational consumers, while the rest 22% is generated from the content providers. ncentives to M ove to NNN While we do know that the BSP prefer s the NNN regime since it provides the flexibility to charge content providers, the surf eit of parameters do not allow us to determine the magnitude of this advantage analytically C omputational results reveal that on average, moving from NN to NNN increases the profit of the BSP by over a third in other words, the BSP will usually have pro nounced
76 incentive to move to the new regime. Thus, the results indicate that under NNN, the BSP lowers the fixed fee to expand the consumer base in order to extract more r ent from one (Outcome 3) or both (Outcome 4) the content providers Although BSP side of the market is lower under NNN than that under NN, the overall profit is higher. I also examine the impact of the different parameters on the BSP Expected ly the BSP valuation The c has two op posing effects when increases the BSP gains as a result of the usage based charge on preferential delivery while the BSP stands to lose due to the increased congestion (and as a result the consumers can be charged less) I find the latter effect dominates the former and the BSP The E ffect of NNN on Market Coverage One of the princ ipal contributions of this chapter is to explore the effect of NN N on market coverage, even as I simu ltaneously modeled the competition between the content providers. In what follows, I denote m arket c overage ( i.e., the total market covered) by and find that, under most cases, In 10873 scenarios out of 10890 total scenarios, we find that market coverage under NNN is higher than NN. In the remaining 17 scenarios (representing about 0.16% of the total) where the market coverage is higher under NN, two common characteristics e merge: First, such a result always happens under Outcome 3, where only content provider G pays the BSP for priority delivery; and second, the ratio is always very high (recall that I had assumed, without loss of generality, that ).
77 The percentage increase of market coverage under NNN: Within the parameter space that I explored, I find that the mean market coverage increases by a little over 7% under NNN as compared to under NN. Impact of different paramet ers on market coverage: I also examine the impact of various parameters on market coverage and find that while, as one might expect, while and have a slight positive impact on market coverage, the del ay cost request arrival rate have a negative effect on total market coverage The Effect on Consumer Surplus NNN regime is to explore how the consumer surplus and social welfare change from those under NN Specifically, if social welfare increases as a result of abandoning net neutrality and more specifically, if the end consumers are better off the idea for the proposed payment mechanisms would gain favor among policymakers. Conversely, if abandoning the principle of net neutrality results in helping just a few private agencies to extract more rent, the idea would find a much less sympathetic audience. The c onsumer surplus under NN is given by the following expression : The consumer surplus under NNN is given by the following expression, depending on whether the final equilibrium is Outcome 3 or Outcome 4 :
78 I find that it is always true that Further, if O utcome 3 is the equilibrium, then it is strictly true that ; i f O utcome 4 is the equilibrium, I sometimes have and in other case s The computational results suggest that the c onsumer surplus for G under NNN is higher than under NN in about 62% of the scenarios and equal to that under NN in the remaining 38% of the scenarios ( but is never lower than that under NN ) Consumer surplus for Y under NNN is lower in about 13% of the scenarios, higher in 49% of the scenarios, and equal to that under NN in the remaining scenarios. Thus the fate of the consumers is slightly nuanced even thou gh consumers as a whole are never worse off (and often better off), with some set of parameter values, the consumers of the less effective revenue generating content provider end up being worse off under NNN as compared to NN. The results also indicate tha t t he increase in consumer surplus ranges from 0% to 62% with a mean increase of about 10.4%. The Effect on Social Welfare S ocial welfare under NN is computed as ( 4 5 ) while the social welfare under NNN is computed as ( 4 6 ) I find that i t is always true that If Outcome 3 is the equilibrium, then the strict inequality holds: ; i f Outcome 4 is th e equilibrium, I find that sometimes and at other time
79 Can NNN R educe C ompetition a mong C ontent P roviders ? T he I ssue of I nnovation at the E dge Since implementing NNN can lead to an equilibrium that in volves Outcome 3, when only G pays for preferential delivery, it is instructive to find out whether such an equilibrium can effectively force content provider Y out of the market (i.e., ). Note that such an outcome would not be p ossible in Outcome 4 neither Y nor G would have any incentive to pay if it had no market share consequently. In 1358 out of the total 10890 (or in about an eighth of the scenarios), I find that to be indeed true. The common characteristics of such outcom es involve a low ( i.e., Y is less effective in generating revenue), and a high consumer request rate for packets and delay cost ( ). When we have a situation where abandoning NN reduces These results play into the hands of observers who have feared that not instituting net neutrality can reduce innovation in the Internet. Fo r example, Vint Cerf, the renowned computer contended that such a payment structure would result in the Internet increasingly resembling mass media, where a few broadband service providers control what the customers effectively may have access to (Waldmeir 2006) Internet start ups often have low revenue generation rates (in many cases they have to spend more than their revenue in order to generate consumer traffic), and resemble the content provider Y with a low Established players (who might resemble content provider G) can preemptively pay the BSP and in some cases (when the packet traffic is relatively high and the consumers have a high disutility associated with delay) shut the new rival out of the market.
80 Fro Established content providers might see an opportunity in paying the priority delivery fees to the BSP and sacrifice profits in the short run in order to preemptively drive out p ossible competition from promising startups in future. While consumers might gain in the short run, with lower access fees (and therefore also support NNN), such a situation might lead to a less innovative Internet, and thereby lessen the experience for al l consumers in the long run. The P ossibility of Free Internet Access to Consumers Under NN, the BSP always charges a positive access fee since that represents its only revenue source However, under NNN, it is possible with some sets of parameter values that the content providers effectively subsidize the consumers for Internet access, i.e., there might be situations where The computational results indicate the circumstances when this is possible. Firs t, the consumer valuation for content h as t o be very low, while delay cost and packet arrival rate are both high. I find that b oth O utcome s 3 and 4 may be the final equilibr ium. When is relatively high compared to O utcome 4 is equilibrium; when is relatively low, O utcome 3 is the equilibrium. The intuition for this type of a scenario is as follo ws: while consumers themselves might not value their Internet access very highly, they are nonetheless profitable to the content providers. Coupled with a high and the consumers would easily drop out of the market if they are charged a positive This would not only deny the BSP from any revenue from the consumers, but would also stop the revenue source from the other side of the platform since the content providers will pa y the priority delivery fee only if there are consumers present on the other side of the platform. Therefore, the BSP decides to drop the access fee for the consumers, in order to retain them, and extract a part of the surplus that the content providers ge t from these consumers.
81 Figure 4 1 Consumers and the market shares of the two content providers Table 4 1. Parameter v alues and s tep s izes V t d Baseline 5 2 1 1 0.5 3.9 0.6 Lower Bound 3 0.5 0.02 0.1 Upper Bound 13 1.5 0.98 3.7 Step Size 1 0.1 0.12 0.4 Total Number 11 11 9 10 Note: The total number of iterations is 11*11*9*10 = 10,890. 0 (Y) 1 (G)
82 CHAPTER 5 CONTENT PROVIDER DIS CRIMINATION AND VERT ICAL INTEGRATION The literature on net neutrality thus far especiall y those which look at the issue from an economic perspective study the problem assuming that the BSP and the content providers are separate entities with conflicting objectives, with the two standing on opposing sides in this debate. However, recent deve lopments indicate that the issue might not be that clearly delineated B roadband service providers like Comcast and AT&T have struck deals with online content providers whereby the latter provide exclusive co branded content through the forme F or example, Comcast and Yahoo! recently signed a multi year agreement so that the broadband services (Shields 2007) Another example is that of AT&T and Yahoo! forming a strategic multi year rel ationship whereby the two companies will offer a co branded version of (BusinessWire 2006) As industry observers have noted, should net neutral ity be abolished, it might motivate BSP to generate their own content (or equivalently, have a strategic relationship with a content provider) and then prioritize delivery of such content to the end consumers. If net neutrality is not enforced, and the BSP their strategic partner, a section of the consumers might switch from their erstwhile content provider to the BSP li ke news, VoIP telephony, music streaming, etc. These developments prompt a new set of questions for the policymaker: How does such vertical integration affect consumer surplus and social welfare? Specifically, under what conditions might the BSP want to ve rtically integrate with a content provider, even though the outcome might result in lower consumer surplus or social welfare?
83 Since the vertically integrated firm can prioritize its own content without paying the priority delivery fees, how does such verti cal integration affect the independent pure play content providers? W ill such an arrangement amount to unfair competition? What are the possible different equilibrium outcomes when the BSP is vertically integrated with a content provider? Will the vertical ly integrated firm prefer no net neutrality over net neutrality? In this chapter I endogenize the BSP net neutrality debate t o answer these questions. In the model, I assume that there is a monopolist BSP who can form a strategic relationship with an online content provider to provide an online service that competes with that from a pure play competitor who only provides online content The latter has to depend on the BSP s content delivered to the end consumers. The Model In this section I set up a game theoretic model to analyze the impact of vertical integration between content provider s and broadband service provider both in the presence and in the absence of net neutra lity. The BSP has two vertical integration related decisions to make: ( 1 ) whether it would like to vertically integrate with one of the two content providers and ( 2 ) if so, which one. As the results for the case of no vertical integration are readily ava il able from Chapter 3 in this dissertation I thus focus the analyses in this chapter on the presupposition that the BSP has vertically integrated with one of the two content providers, a nd then compare and contrast these findings with those of Chapter 3 to address the BSP Hence, t here are three types of players in the game the vertically integrated monopolist BSP the two content providers (one of whom is vertically integrated with the BSP ) and the end consumers. The BSP se rves as an intermediary and transmits content from the content providers to end consumers. Since the BSP is vertically integrated, it has its own content (from its strategic
84 partner) that competes with the other independent pure play content provider ( I ca ll the independent content provider C) for the attention of the end consumers. Content Providers To model the competition between the content providers, I assume two content providers Y and G who offer their contents for free to the end consumers. (One of the two content providers is vertically integrated with the BSP .) I assume that the content from these two providers are horizontally differentiated in a Hotelling sense with the two of them located on the two ends of line segment A s before I assume that the content providers provide their Let and be the average revenue generated per packet requested by the end consumer, where I denote the r evenue generation rate of the vertically integrated BSP by and the revenue generation rate of the independent content provider C, where by Thus, and if the BSP vertically integrates with G and and if the BSP vertically integrates with Y Therefore, depending on the BSP can be either greater than or less than The competition between content provider C and the BSP is driven by the fact that a larger consumer base will lead to greater advertising revenue. Vertically Integrated B SP Following the central tenets of the model in Chapter 3 I assume a monopolist BSP provides Internet access as well as its own content to the end consumers. The BSP can provide its content as the result of a vertical integration with a content provider the exact mechanism
85 for this integration might be achieved through an outright merger between the two firms, or thro ugh a strategic alliance. 9 Figure 5 1 shows a schematic of the market structure of the model. The BSP charges the consumers a fixed fee for Internet access and potentially charges content provider C a per packet fee enforced. Consumers Consumers request content from either the vertically integrated BSP or content provider C. Similar to the models in Chapters 3 and 4, f or an arbitrary consumer the fit cost associated with deviation from his ideal content is if the consumer chooses the BSP and if the consumer chooses content from provider C. Then the utility function for the BSP ( 5 1 ) ( 5 2 ) In the absence of net neutrality, I use a two class priority queue to model the BSP transmission service. The detailed congestion cost and the corresponding utility functions for consumers under no net neutrality will be discusse d in the section of no net neutrality I define two indicator functions as follows to represent whether the BSP would prioritize its own content and whether content provider C would pay for the preferential delivery. 9 The details of the profit sharing agreement between the two firms in a strategic alliance are not germane to this
86 ( 5 3 ) ( 5 4 ) The timing of the four stage game is as shown in Figure 5 2 In stage 1, the BSP decides ( 1 ) whether to integrate with a content provider and ( 2 ) if so, integrate with who m. In stage 2, the broadband provider announces and In stage 3, the BSP decides whether to give its own content preferential treatment and prefe rential delivery. In stage 4, consumers choose content from either the BSP or content provider C. Given this particular timing, the strategy of the BSP is to calculate its maximum profit under different scenarios, and then, depending on the underlying para meter values (see Appendix A for a list of the various parameters and variables used in the text), choose whether to observe net neutrality, or if the regulatory environment allows him to do so choose to prioritize its own content and (for a fee) the c ontent from provider C to the end consumers. Thus, the monopolist BSP can calculate what would be the profit under different scenarios and then choose that particular pricing strategy that will maximize its profit for a given set of parameter values. In th e following two sections, I analyze these different scenarios by solv ing the BSP profit maximization problems under different regulatory regimes (net neutrality and no net neutrality) and pricing arrangements. Note that there is only one pricing decision under net neutrality (the access fee that the BSP charges the consumers), while under no net neutrality, the BSP can ch arge both the consumers and content provider C (charging the latter for the service that prioritizes the delivery of its content to the end consumers).
87 Net Neutrality With net neutrality in place, the BSP is forbidden from providing the service of and charging for preferential delivery of data packets. Then the marginal consumer who is indifferent from the BSP and content provider C can be determined by i.e., ( 5 5 ) which implies Therefore the demand for the BSP and content provider C are both The resulting Internet access fee is The BSP Internet access charge from consumers i.e., and advertisement revenues from advertisers i.e., No Net Neutrality Without net neutrality, the BSP has the option to provide a preferential delivery for data packets from content provider C a nd the BSP itself. Depending on whether the BSP prioritizes its own content and whether C pays for the preferential delivery, there are four different outc omes as follows. Outcome 1 : C does not pay for priority delivery of its own content, and content from n either provider is prioritized ; Outcome 2 : the BSP prioritizes its own content at the expense of that of C who does not pay the priority delivery fee; Outcome 3 : the BSP ized at the expense of its own ; and Outcome 4 : C pays the BSP so that the BSP does not prioritize its own content over that of C (in other words, content from both providers receive the same priority ). Note that all these four outcomes are BSP for example, if it is charging C for priority delivery, it can
88 create a contract that specifies whether it will in turn not prioritize its own content ( i.e., Outcome a priori none of the outcomes can be ruled out, since based on the values of the different parameters, any one of them might gener ate the highest profit for the B SP. Outc ome 1: The B SP does not prioritize its own content and content provider C does not pay for the preferential delivery ( ). Outcome 1 is essentially the equilibrium equivalent to net neutrality. The corresponding marginal consumer is determined by i.e., ( 5 6 ) which implies Under this scenario, the BSP ( 5 7 ) where constraints and constraints and are incentive compatibility cons traints for the B SP and content provider C. Note that for its incentive compatibility considerations, the B SP compares its profit under Outcome 1 with that under Outcome 2 (where it prioritizes its own content over that of C). To consider the incentive com patibility constraint of content provider C, meanwhile, the BSP
89 profit under Outcome 1 to that under Outcome 3 of its own. Some algebra shows that regardless of parameter values, the incentive compatibility constraint for the BSP can never be satisfied. In other words, if content provider C does not pay for the preferential delivery, the BSP is always better off prioritiz ing its own content (see Table 5 1 for a comparison of the BSP profits under the different scenarios). Therefore Outcome 1 is not an equilibrium, unless the BSP is prohibited through regulation from differentially prioritizing its own content. Outcome 2: Under Outcome 2, the B SP prioritize s its own content and conte nt provider C does not pay for the preferential delivery ( ) In this case, data packets provided by the B SP and content provider C are transmitted with different priorities by the B SP. I use a two clas s priority queue with preemption model to depict the waiting ti me. For a given consumer d ata packets from the B SP are transmitted with higher priority whose waiting time is while data packets from con tent provider C are transmitted with lower priority whose waiting time is Correspondingly the marginal consumer is determined by i.e., ( 5 8 ) which leads to a higher market share for the B SP ( ) than content provider C ( ).
90 The B SP maximizes its profit by solving ( 5 9 ) where, just as in Formulation ( 5 7 ) constraints and constraints and constraints and are incentive compatibility constraints for the B SP and content provider C. Note that if Outcome 2 holds, then the B SP does not generate any revenue from priority delivery, since the only content tha t is prioritized is from its strategic partner. In BSP will BSP (so that its content gets lower priority for delivery than the BSP Outcome 4, when C pays to ensure that the delivery of its content will not be relatively degraded with respect to the BSP Outcome 2 can only arise in equi librium if i.e., the B SP will offer to zero, the relative profitability of the two content providers can be measured from their revenues). If however 10 Outcome 2 is dominated by other scenarios. 10 I assume that when the BSP is indifferent between two different outcomes (i.e., between Outcome 2 and Outcome 4, or between Outcome 3 and Outcome 4), it chooses Outcome 4.
91 Outcome 3: The BSP does not prioritize its own content and content pr ovider C pays for the preferential delivery ( ). As opposed to Outcome 2, for a given consumer in Outcome 3 it is the data packets from content provider C that are transmitte d with higher priority with a waiting time of while data packets from the BSP are transmitted with lower priority with waiting time of Correspondingly the marginal consumer under Outcome 3, is determined by i.e., ( 5 10 ) which leads t o a lower market share for the B SP ( ) than content provider C ( ).The B SP maximize its profit by solving ( 5 11 ) Now, in order to satisfy its own incentive compatibility constraints, the BSP has to make sure that its profits under Outcome 3, when C pays and the d elivery of its own content is degraded with respect to that of C, is at least as high as under Outcome 4, when C pays and the content from both providers receive equal priority. Note that depending on the relative
92 magnitudes of the average per consumer adv ertising revenue that is generated by the BSP and C, the BSP will sometimes willingly degrade delivery of its own content (Outcome 3) in order to extract the surplus from the advertising revenue that C generates from its bigger market share. The additional surplus extracted from C might be enough to more than compensate the loss of compatibility constraints, the BSP has to ensure that by paying and getting prior ity delivery of its packets, C profit is at least as high as under Outcome 1, when it does not pay and the delivery of its content does not receive priority (but is not degraded either). A nalogous to Outcome 2, Outcome 3 can only arise in equilibrium if i.e. the BSP will offer this pricing policy only if it is the Outcome 3 is dominated by other scenarios. We also note that since the in dependent content provider is more effective in generating revenue, the BSP can credibly enforce this pricing/prioritization strategy where content provider C has reasons to believe that the BSP will t he market share for content provider C will increase and therefore the BSP can extract more surplus from content provider C through charges for priority service Outcome 4: T he BSP prioritizes its own content and content provider C pays for the preferentia l delivery ( ) to ensure that its own packets do not get degraded with respect to those of the BSP as in Outcome 2. In Outcome 4, data packets from both the B delivery in other words, d elivery of neither content is degraded with respect to the other.
93 Therefore, packets from each provider face the same waiting time So the marginal consumer is determined by i.e., ( 5 12 ) which leads to The BSP maximization problem under Outcome 4 is: ( 5 13 ) For the BSP the scenar io is the reverse of Outcome 3, and hence, to maintain incentive compatibility, the BSP has to ensure that its profit under Outcome 4 is at least as high as under Outcome 3. For content provider C, there remains an incentive to pay the BSP only if by doing so (and thereby not have the delivery of its content degraded with respect to those of the BSP as in Outcome 2), it can generate profits that are as high as under Outcome 2. Proposition 5 1 summarizes the results of equilibrium of the game. Proposition 5 1 : (Equilibrium of the game with a vertically integrated BSP ) There are three possible equilibria of the game. Under Case A : Outcome 2 is the equilibrium Under Case B : Outc ome 3 is the equilibrium.
94 U nder Case C : Outcome 4 is the equilibrium. Case C can be further divided into two sub cases: Under Case C1 : ; U nder Case C2 : Proof: See Appendix B. Figure 5 3 shows graphically the effect of the relative magnitudes of and on the final equilibrium. One can think of the graph being divided into two areas by the line with the bottom half corresponding to the case when the vertically integrated B SP is more profitable in generating advertising revenue than its pure play competitor, while the t op half represents the opposite scenario. The figure can be divided into three main regions, which I denote as Case A, B and C respectively, and Case C can be further divided into two regions C1 and C2. The shaded region in the bottom half of the quadrant represents Case A, or Outcome 2. The other shaded region, which is in the top half of the quadrant represents Case B, or Outcome 3. The remaining area represents Case C, or Outcome 4. With respect to the regions outlined in Figure 5 3 Outcome 4 is the equ ilibrium in the region in between that of Outcome 2 and Outcome 3. This is where the profitability of the BSP N ote that the fixed fee is higher in Outcome
95 4 than in Outcomes 2 and 3, and the BSP now gets relatively more of its profit from the consumers than from the content providers as compared to Outcomes 2 and 3. As noted earlier, Case A or Case B occurs only when the relative revenue generation rates a nd a necessary is given by the straight lines that demarcate the different regions). When the revenue dominates. In this scenario, the BSP can credibly ask for compensation from its competitor content provider as an assurance for not prioritizing its (the BSP compare this outcome with th at of Case B. In the latter case, the BSP which is relatively much less profitable in generating revenue from its own content than its competitor, finds it to its en extract the su rplus from the latter which has a higher market share, than to give its content the same priority as that of its competitor (which it does under Case C, when the BSP is relatively more profitable). This in the sabota ge literature that a vertically integrated (Mandy 2000) The two regions in Case C, C1 and C2, warrant a separate explanation. The level of the priority access fee is de termined by two constraints: (1) it has to be low enough to attract content provider C to pay and (2) as opposed to Case B discussed above, has to be low enough so that the BSP does not have the incentive to degrade the delivery of its own content. Case C1 denotes the region where constraint 1 binds, while Case C2 denotes the region where constraint 2 binds. I end this section with a discussion of the implications of Case A and B. Under Case A, the BSP prioritizes its own content at the expense of its competitor, who prefers not to pay for that
96 privilege. The BSP thus gets an advantage over its competitor without the requirement of a comparable rent ( since its own content ) and this is something that should definitely be of interest to a policymaker deliberating on this issue since such an arrangement might be tantamount to unfair competition Under Case B, we have an outcome which looks counter intuitive at first sight: the BSP deliberately de prioritizes its own content with respect to its competitor in exchange for a fee. As the discussion in the previous paragraph indicates, this outcome is plausible : the BSP gains more from the fee from its competitor (who has a higher m arket share) than it would have generated from the advertising revenue of its own content ( and commanding the same market share) by delivering its own content with the same priority as that of its competitor. The BSP Facing two content providers Y and G (with respective revenue rate and where ), the BSP needs to examine the area above the line of Figure 5 4 a in arrivi ng its vertical integration decision. F or an arbitrary point on the p arameter space in Figure 5 4 a the BSP is one of the following: Y in Figure 5 4 b) G in Figure 5 4 b) Figure 5 4 a shows the various regions which we need to analyze in order to exhaustively determine the BSP These are best explained using Table 5 1. The expressions in the cells of the table refer to the profit of the BSP under various conditions. The subscript VI (short for vertical integration) refers to the BSP nt pro viders. Further, the subscript refers to the BSP Y and refers
97 to the BSP G The subscript NVI (short for no vertical integration) refer s to the BSP 2, 3, 41 and 42 refer to the four potential equilibria: 2 corresponds to Case A of Proposition 5 1 where Outcome 2 is the equilibrium, 3 corresponds to Case B of P roposition 5 1 where Outcome 3 is the equilibrium, 41 corresponds to Case C1 of Proposition 5 1 where Outcome 4 is the equilibrium and 42 corresponds to Case C2 of Proposition 5 1 where Outcome 4 is the equilibrium. Note that when there is no vertical inte gration, there are no equivalent cases that correspond to Case C1 and C2 of the vertical integration setting. However, to properly illustrate the region corresponding to Outcome 4 under no vertical integration, we need to include the straight line that divides regions 3 and 6. This line separates the two cases where Outcome 3 is the equilibrium and where Outcome 4 is the equilibrium (see Chapter 3 for details) Thus, region 1 represents the space where O utcome 3 is the equilibrium if the BSP integrates either with content provider Y or neither of them, and Outcome 2 is the equilibrium if the BSP integrates with content provider G The other regions can be interpreted similarly. By comparing the profits of the BSP for the three different choices in each of these regions, I find that the BSP will always prefer to integrate with content provider G (the cells with stars in Table 5 1 represent the BSP sho wed in Appendix B These results however do not tell us whether the social planner would also prefer such an outcome. That discussion is provided later in the section of welfare comparison, where I consider the choices of the social planner and compare the m with the BSP choices.
98 The Welfare Effect of Net Neutrality The different scenarios in the previo us section, one under a net neutrality regime (or NN for short) and four under no net neutrality (NNN) list out exhaustively the various pricing/prioritiza tion strategies of the BSP Depending on the specific values of the parameters, the BSP can calculate the exact values of the optimum profit under each scenario, and deci de on the best pricing strategy and correspondingly We have already observed that given the choice the BSP will never opt for abiding by the principles of net neutrality in other words, unless enforced, net neutrality will never be a natural outcome of the game. It is imperative th erefore from a regulatory perspective to find out whether it is in the best interests of the other players (the content providers and the consumers), as well as in terms of the total social welfare, to regulate enforcement of network neutrality. This secti on is devoted to that discussion I discuss the impact of net neutrality (or its abolishment) on the payoffs to the three players and the overall welfare. Proposition 5 2 summarizes the findings. Proposition 5 2 : (The welfare impact of net neutrality with a vertically integrated BSP ) Under Case A where and Case B where both consumer surplus and social welfare are higher under NNN than under NN Under Case C where both consum er surplus and social welfare are the same under NN and NNN Proof: See Appendix B
99 The expressions of consumer surplus and social welfare are provided in Table 5 2 which shows the comparison of the prices charged by the BSP the profits of the BSP and th at of the content provider C, the consumer surplus and the total social welfare under NN and in the three potential equilibria under NNN which I denoted earlier as Cases A, B and C respectively Case C is in turn divided into two subcases C1 and C2, as di scussed in the previous section. Table 5 2 also shows the result of comparing the magnitude of these output variables under the different scenarios of NNN to that under NN, and in all the cases, we establish how the expressions compare regardless of input parameter values. As can be seen from Table 5 2 with NNN than with NN under Cases A and B and unchanged under Case C. The corresponding consumer surpluses are higher under Cases A and B as compared to NN, and unchanged under Case C. Similarly, the total social welfare with vertical integration increases with NNN under Cases A and B as compared to NN, w hile it is unchanged in Case C. The vertically integrated BSP profit is always higher under NNN, and therefore the BSP will always prefer NNN over NN. The expressions in Table 5 2 establish that the content provider C has either the same profit (Case B) or is worse off under NNN. Interestingly, content provider C has the same profit in Case B that it has under NN, even th ough it has a higher market share than under NN this is so because the BSP is able to extract the additional surplus from C completely through the priority access fee. In Case A, content provider C has a lower market share than under NN, and consequently a lower profit In Case C, content provider C and the BSP have the same market lower than under NN.
100 Welfare Comparison with Vertical Integration (VI) and No Vertical Integration (NVI) T he issue of vertical integration in the online environment whereby a BSP is both a provider and the delivery agent of the content is especially significant from the perspective of the debate over net neutrality. C ommentator s of net neutrality have argued that the vertically integrated firm has an unfair competitive advantage over the pure play content provider since the former can effectively act as a gatekeeper who can control the experience of the end consumers when they consume the content. Many consumers might find the delay of accessing particular websites having a disutility high enough to switch over to a competing provider, and since the vertically integrated firm can ask for compensation for not de prioritizing the content of the rival content provider (or simply prioritize its own content without paying a fee), it can leave the pure play content provider at a competitive disadvantage. This insalubrious effect on the competition can be more palatable if it is found t hat the total social surplus as a result of the vertical integration is higher than what it would be if the BSP were not vertically integrated. In the latter scenario (i.e. with no vertical integration) the two competing content providers would have to p ay the BSP in order to have their respective content delivered with priority to the end consumers. The various scenarios under this setting were analyzed in Chapter 3 We now proceed to compare the results in Chapter 3 with the findings in this chapter Fi gure 5 5 visually summarizes the comparisons. Proposition 5 3 : (The welfare effect of vertical integration) When social welfare is lower under VI than under NVI
101 When social welfare is higher under VI than under NVI Otherwise, social welfare remains unchanged. Proof: See Appendix B As can be seen from the graph in Figure 5 5 the crucial factors that determine the comparison of the social welfare with vertical integration (VI) as to when there is no vertical integration (NVI) are the relative magnitudes of the advertisement generation capability of the two content providers (in other words, the ratio of to ), and whether the BSP has a strategic relat ionship with the more profitable content provider. In the graphs, the shaded regions represent the areas where the surpluses with VI and with NVI differ from one another. The term signifies the difference between total social we lfare under VI and under NVI. If the BSP has its strategic relationship with the less profitable content provider, the applicable region is the upper half of the quadrant that plots to I n that situatio n, the surplus with VI is never greater than the surplus with NVI, and in fact, within the narrow strip of the shaded region i.e. the surplus is lower. Conversely, if the BSP has its strategic relationship with the more profit able content provider 11 the total social welfare with VI is never lower than with NVI, and for the relatively thick shaded strip shown in Figure 5 5 the total surplus with vertical integration is actually higher than it would be with no vertical integrati on 12 11 The corresponding region in Figure 5 5 is the lower half of the quadrant where 12 Simple algebra shows that lower strip is relatively thicker compared to the upper strip.
102 The reason for this can be understood as follows: In the region of the upper shaded strip, with no vertical integration, only content provider C pays and has its content prioritized at the expense of the content from the other content provider. Howeve r, with vertical integration, while content provider C is still the only paying content provider, in the shaded region, the content from the other content provider (now the BSP ) receives equal priority. The corresponding loss of surplus of C is not fully c ompensated for by the gain of the BSP from its content, since the latter has a less effective revenue generation rate. The explanation for the lower shaded region is analogous: with no vertical integration, in this region, both content providers would have paid for equal priority of delivery, but with vertical integration, neither content providers pay ; however with the content of the more effective content provider (now the BSP ) being delivered with priority, the loss of C is more than compensated for by t he gain of the BSP From the perspective of a policymaker, the results of this analysis are extremely significant. If net neutrality is not enforced, vertical integration can increase social welfare when the vertically integrated firm is also the more pro fitable co ntent provider. However, the discussion might not always be this straightforward: one can certainly think of a scenario where the currently dominant content provider, envisaging future competition from a promising rival upstart (who right now lac ks the financial prowess to generate more revenue from its own content), can preemptively merge with the BSP and put the rival startup at a disadvantage, and this might lead to a less socially beneficial outcome in the long run. However if the vertically integrated firm is the less profitable content provider (e.g. when the local BSP of parameter values, social welfare can decrease (and will never increase), and therefore t he policymaker will have reasons to subject such a service to more scrutiny.
103 Table 5 3 summarizes the preferences of the policymaker for vertical integration. As in Table 5 2, we need to consider the social welfare in all the seven regions. In regions 2 an d 3 (corresponds to the upper shaded strip of Figure 5 5 ), the social planner prefers no vertical integration. In regions 4 and 6 (corresponds to the lower shaded strip of Figure 5 5 ), the social planner prefers vertical integration of the BSP and content provider G In the other regions, the social planner has no preference between vertical integration and no vertical integration. The cells with stars in Table 5 3 represent the s preferences. I summarize the comparison results in Proposition 5 4. Proposition 5 4 : (The vertical integration compared to the BSP ) Comparing the BSP following: When the BSP ical integration integration while the BSP prefers to integrate with content provider G When the BSP is BSP prefer to integrate with content provider G Proof: See Appendix B
104 Figure 5 1. Market s tructure with v ertical i ntegration of c ontent and b roadband s ervices Fig ure 5 2 The sequence of events in the game with vertical integration The BSP announces and Stage 2 The BSP decides whether to give its own content preferential treatment Stage 3 Consumers choose content from the BSP or content provider C Stage 4 The BSP decides (1) whether to integrate and (2) if so, integrate with whom Stage 1 Advertisers Advertisers Content Provider C BSP Consumers
105 Figure 5 3 The e quilibria with v ertical i ntegration Case C1 Case C 0 Case A Case B Case C2
106 A B Figure 5 4 The BSP v ertical i ntegration d ecision A) The B space integration with G or Y. 4 5 7 0 1 3 2 6 0
107 Figure 5 5 The e ffect of v ertical i ntegration on s ocial w elfare Table 5 1. The BSP vertical integration choices integrate wit h Y integrate with G integrate with neither Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7 0
108 Table 5 2 Comparison of various economic outcomes of interest under NN and NNN Note: The text in parenthesis shows how those economic outcomes change when moving from NN to NNN) NN (Benchmark) NNN (Case A: Outcome 2 content is prioritized) NNN (Case B: Outcome 3 NNN (Case C: Outcome 4 Both the BSP and equally prioritized) Case C1 CaseC2 (Lower) ( Lower ) ( Unchanged ) ( Unchanged ) N/A N/A (Better off) ( Better off ) ( Be tter off ) ( Better off ) Content Profit (Worse off) ( Unchanged ) ( Worse off ) ( Worse off ) Consumer Surplus ( Better off ) ( Better off ) ( Unchanged ) ( Unchanged ) Social Welfare ( Increased ) ( Increased ) ( Unchanged ) ( Unchanged )
109 Ta ble 5 3. integrate with Y integrate with G integrate with neither Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7
110 CHAPTER 6 USER DISCRIMINATION I n 2007, it was independently verified that the broadband Internet service provide r Comcast was slowing down network traffic within its servers that originated from the popular peer to peer (P2P) networks (McCullagh 2007) After initially denying any such behavior, Comcast defended its actions by claiming that th e traffic from the P2P networks, which was dominated by just a small fraction of the total number of users, was slowing down the network traffic for the rest of the users The United States Federal Communications Commission (FCC) providing furthe debate So far, the growing literature that has analyzed these economic issues of net neutrality has modeled that interested party as the content providers who are jockeying for a position in the the aforementioned Comcast example shows, the interested party might well be some of the consumers themselves who are willing to pay a fee to have their requested packets delivered with priority. In other words, a data packet traveling from its origin to of its journey either side, whereby the BSP charges content providers for preferential delive ry of their dem side, whereby the BSP charges the individual consumers a fee for the priority delivery of their requested content (or equivalently, de prioritizes the requested cont ent in the absence of the fee). In this chapter, the focus is on the latter aspect of net neutrality, whereby I analyze the e conomic rationale for and against the proposals put forth by several broadband service providers who intend to differentiate between different classes of users. For example, the cable broadband service provider Time Warner Cable has recently started an exp eriment in certain
111 markets where they plan to charge Internet customers based on how much Web data they consume. The experiment started in a single market (Beaumont, TX) in the summer of 2008, and the company plans to introduce tiered pricing in several ot her markets in the near future. By charging a premium to the heaviest broadband users, much the same way cell phone providers collect fees from subscribers who exceed their allotted minutes, Time Warner would upend a longstanding uniform pricing strategy a mong (fixed line) Internet service providers in the United States, whereby phone and cable companies have charged flat fees for unlimited access to the Web. AT&T has started a similar experiment with its own customers, also in Beaumont, TX. As expected, su ch experiments have reignited the net neutrality debate. Proponents of net neutrality consumer advocates and online content providers, for example have opined that that a tiered Web use pricing would limit customer choice and could stifle innovation by crimping demand for high bandwidth services such as online video and music (Al Chalabi 2008) However, cable and phone c ompanies have countered by saying that they need the flexibility in setting prices for use of large, expensive, heavily used broadband networks, so as to effectively serve the majority of their customers and encourage greater efficiency in the way customer s use capacity (Tweney 2008) As consumers spend more time online, and also use the Internet to consume various types of data intensive content (like music and video a high definition movie typically consumes around 8 GB of traffic), the decision to charge data consumption by volume can be expected to have profound implications in the way online content is consumed in future. In such scenarios, e are actually not downloading that much data. The company's trial in Beaumont, TX, lasted several months: of the
112 10,000 broadband customers enrolled which represented about 25% of the company's total number of consumers in Beaumont about 14% exceeded their cap and had to pay additional fees that averaged about $19 a month. Time Warner Cable also discovered that the top 25% of users consumed 100 times more data than the bottom 25% of users, suggesting an enormous gap in usage patterns. Broadband service providers have often mentioned that as more and more people download TV shows and movies, particularly those in high definition, the broadband network infrastructure face s enormous strain. Time Warner Cable has said its strategy is intended to alleviate s ome of that strain, with users self regulating themselves under the new plan. But critics have expressed concerns that the pricing scheme will discourage broadband use and impede new online media businesses before they even have a chance to flourish. The e ntire debate has raised a number of unanswered questions that are of interest to researchers and practitioners alike, not to mention the regulatory agencies. While legal scholars might debate whether such pricing plans (as those that Time Warner and AT&T a re experimenting with) or prioritization strategies (as Comcast briefly attempted) are fair on the consumers, it is an entirely separate issue whether there are economic incentives for the BSPs to pursue such strategies that go against the net neutrality p rinciple. In other words, facing a highly dynamic and differentiated data usage patterns from different classes of users, would a BSP gain (as compared to the status quo) by employing different pricing and/or packet prioritization stra tegies? In the first part of the analysis, I explore this issue. While the BSP might prefer not to adhere to the principles of net neutrality under certain circumstances, such a move might be detrimental to the consumers or the society as a whole. perspective, the issue is somewhat different: would the
113 abolis side result in lower consumer surplus or social welfare? Depending on that answer, the social planner might wish to regulate on the issue. In this chapte r I explore these issues and model them in an analytical framework and examine the economic impacts of user discrimination and net neutrality from the perspectives of both th e BSP and the social planner. I characterize the dynamic and differentiated data demand of the end consumers by their valuations for data consumption and their usage patterns. Under the current scenario (which can be thought of as the net neutrality model with a uniform fixed fee pricing strategy), all users are charged the same fixed price for accessing broadband content. Both types of users face similar delays while accessing their desired content the delay arises fixed capacity. Facing th is heterogeneous user data demand, broadband service providers have two potential instruments for user discrimination price discrimination and traffic prioritization. If the BSP is allowed to differentially charge its users and/or priori tize their reques ted content, I explore six different strategies that it might employ: Broadband user traffic from different user types face the same delay, and all users are charged the same fixed fee (i.e. the status quo). Broadband user traffic from different user type s face the same delay, and different types of users are charged different fixed fees. Broadband user traffic from different user types face the same delay, and different types of users are charged a two part tariff. Broadband user traffic from different us er types face different delays, and all users are charged the same fixed fee. Broadband user traffic from different user types face different delays, and different types of users are charged different fixed fees. Broadband user traffic from different user types face different delays, and different types of users are charged a two part tariff.
114 The first three options (where all the broadband users face the same delay for their packets) cover different pricing strategies under net neutrality (or NN for short) while the last three options cover the different pricing strategies under no net neutrality (NNN). Another way to look at these options would be to think of the first three as represent ing strategies that the BSP can adopt if it is allowed to use only pr ice discrimination, while the last three would represent strategies where the BSP is allowed to use both price and traffic prioritization as discriminating tools. These six different options help us model a broad swath of strategies that a BSP might employ valuations for content and their usage patterns, different types of pricing and traffic prioritization regimes yield different profits for the BSP. However the optimal choice for the BSP might not coincide with that of a policymaker who intends to maximize the total social surplus. The results of the analysis should therefore be useful both for the broadband service providers as they mull over the introduction of the dif ferent pricing/prioritization strategies in an age where consumers increasingly get their information and entertainment online, and for policymakers who might T he Model As before, I assume a monopolist BSP who provides Internet access to the end consumers. To model the demand for broad band Internet access service, I consider a unit mass of end con sumers. As mentioned earlier, I assume that there are two types of users: a fraction of H type consumers and fra ction of L type consumers. High type users request more content (the requested rate of data packets by the two user types are given by and respectively, where ) and have higher valuation for that content ( ) than the Low type users. arge a uniform
115 fixed fee ( ) per unit time to all consumers, different fixed fees ( and ) per unit time to different types of consumers, or a usage based fee ( ) per packet to consumers for Internet access, a pricing strategy that has been already employed in some Scandinavian countries (Bandyopadhyay and Cheng 2006; Economist 2003) Since the consumers are serviced by the BSP which has a fixed network i nfrastructure capacity, the former encounter a disutility while ( 6 1 ) where and is the delay for type consumers. Consumers request data from va rious websites and the requested data packets are I assume Poisson process with arrival rate and for H type and L typ e consumers respectively. The gross valuations the two types of consumers receive are denoted by and The delay for the consumers under net neutrality is: ( 6 2 ) In the absence of net neutrality, the BSP may prioritize data traffic based on user types. In this context, we no te that the technology to discriminate packets and streamline Internet traffic has been a vailable at minimal cost, and I therefore assume that there is no additional expense incurred by the BSP to implement a mechanism that enables preferential delivery of content. I use a two type and L type consumers receive the same priority for their traffic, then the congestion cost and the corresponding utility functi on would remain the same as in E quation ( 6 2 ) However, i f H type
116 consumers receive higher priority while L type consumers receive lower priority, then the delay costs f or the two typ es are given in ( 6 3 ) ( 6 3 ) On the other hand, if L type consumers receive higher priority while H type consumers receive lower priority, then the delay for the two types are given in ( 6 4 ) ( 6 4 ) In terms of pricing, the BSP may charge a uniform fixed fee to all consumers or charge different fixed fees to different types of consumers for Internet access. The potential regulation of net neutrality limits the BSP from selectively prioritizing the Internet traffic. In th e absence of net neutrality, the BSP can also discriminate against different types of consumers through traffic prioritization. In the next two sections, I model these scenarios. Net Neutrality In this section, I analyze three potential pricing structures for the BSP under net neutrality uniform fixed fee, differential fixed fees and charging a two part tariff. O ption NN1: Uniform F ixed F ee U nder N et N eutrality Under net neutrality all consumers receive the same priority and therefore face the same conges tion for data transmission. The most simple and common pricing mechanism for the BSP is to charge problem is formulated in ( 6 5 )
117 ( 6 5 ) Constraint is the participation constraint for H type consumers and constraint is the participation constraint for L type consumers. Since the BSP will charge a fixed access fee that is high enough to just keep the L type consumers to participate, i.e., and the BSP then receives a corresponding profit of The corresponding consumer surplus, defined as the sum of the utility of all consumers, is given by consumer surplus, is Option NN2: Differential F ixed F ees U nder N et N eutrality It is easy to see that this option reduces to the option NN1 above. This is because the BSP has just one service offering at its disposal, and therefore will not be able to differentiate between the two classes of users by using different prices (if the two user types are offered tw o different price points, the H type users will always choose the lower price, as would the L type users). The formal statement
118 ( 6 6 ) Constraints and are participation constraints for H type and L type consumers resp ectively. Constraints and are incentive compatibility constraints for H type and L type consumers respectively. Constraint can be reduced to and Constraint can be reduced to So As a result, under net neutrality Option NN2 can be reduced to Option NN1 with The corresponding consumer surplu s will still be and the social welfare will be given by Option NN3: Two P art T ariff U nder N et N eutrality Under this option, the BSP charges a two part tariff for Internet access a lump sum fee and a per unit charge
119 ( 6 7 ) Constraint is the participation constraint for H type consumers and constraint is the participation constraint for L B for derivation details), I a consumption are comparable (I denote this as Case NN3_1, with th e exact criterion being ), the BSP will charge a positive lump sum fee and a positive usage based fee ; however, if the two types of consumers differ significantly in their val uations for their requested content (or more precisely, if which I denote as Case NN3_2), the BSP will charge a zero lump sum fee and rely only on usage based fee: and The corresponding consumer surpluses are and The resulting social welfare is Note that under Case NN3_1, the entire consumer surplus is extracted away completely by the BSP.
120 No Net Neutrality In this section, I fixed fees and two part tariff) under NNN. In the absence of any net neutrality regulation, broadband service providers have one extra set of instru ments to discriminate between end users: the BSP may assign different priorities to different types of traffic. Technically, BSPs first identify the data destination by inspecting data packets transmitted through the network. The BSPs then either charge th e same access fee for all consumers and downgrade data transmission for heav y users (H type consumers in the model) or charge a higher access fee and then assign a higher priority for the data packets requested by H type consumers. Just as I analyzed the pricing strategies under NN, I now look into the three analogous pricing regimes under NNN. Option NNN1: Uniform F ixed F ee U nder N o N et N eutrality When the BSP charge a uniform price to both types of the consumers, it has the incentive to assign a lower pr iority to data packets from H type users because of their heavy use of the ( 6 8 ) Constraint is the participation constraint for H type consumers (reflecting their higher wait times in the prioritized queue) and constraint is the participation constraint for L type consumers. Notice that I assume to ensure that this
121 scenario is feasible. Both constraints give upper bounds for the access fee One can deri ve the equilibrium by comparing the two upper bounds. Case NNN1_1: If i.e., then and The corresponding consumer surplus is The expression for social welfare is Case NNN1_2: If then and The corresponding consumer surplus is
122 and the corresponding social welfare is given by the following expression: Option NNN2: Differential F ixed F ees U nder N o N et N eutrality Under this option, the BSP charges a higher pr ice for higher quality of the data transmission service through the Internet. Specifically, the BSP would offer the Internet access service with congestion cost at a fixed price to H type consumers and the Internet access service with congestion cost at a fixed price to L
123 ( 6 9 ) Constraint is the participation constraint for H type consumers and constraint is the participation constraint for L type consumers. Notice that I assume to ensure the feasibility of this outcome Constraints and are incentive compatibili ty constraints for H type and L type cons umers respectively. Constraint can be reduced to Constraint can be reduced to Therefore, From constraint we get
124 From constraint we get Since and The corresponding consumer surplus is and the social welfare is given by Option NNN3: Two P art T ariff U nder N o N et N eutrality Und er this scenario, the BSP charges the H type consumers a two part tariff to ensure a preferential delivery of their data packets, while the L type consumers are charged only a lump decision problem can be formulated as:
125 ( 6 10 ) Constraint is the participation constraint for H type consum ers and constraint is the participation constraint for L type consumers. Constraints and are incentive compati bility constraints for H type and L type cons umers respectively. Constraint can be reduced to Con straint can be reduced to So S ubstituting back to constraint gives Constraint implies
126 So and The correspondin g consumer surplus is Therefore the social welfare is the BSP has six options and I summarize these six o ptions in the Table 6 1 In the previous two sections, I (under NNN) different priorities for different user classes. In the next section, I compare these different options, and thus explore the con ditions under which the BSP might choose any one of them. In this section I profit. P reference for P ricing S tructure U nder N et N eutrality ( C hoice A mo ng T hree O ptions NN1, NN2, NN3) Under net neutrality, the BSP is limited to just the pricing mechanisms to discriminate
127 NN3) under net neutrality yields and This result is summarized in P roposition 6 1 Proposition 6 1: ( The Under net neutrality, the BSP prefers a two part tariff. Proof : S ee Appendix B. This r esult is intuitive since a two part tariff pr ovides two pricing instruments (fixed fee and usage based fee) and therefore can extract more consumer surplus compared to the uniform fixed fee only for the BSP reference for P ricing S tructure U nder N o N et N eutrality ( C hoice A mong T hree O ptions NNN1, NNN2, NNN3) In the absence of net neutrality, the BSP may either charge the same price to both types of consumers and set a lower priority to data packets from H type consumers (NNN1), or charge a higher price and set a higher priority to H type consumers (NNN2), or charge H type consumers a usage based fee to get a higher priority (NNN3). The first option yields profit levels if or if The second and third option generates the same profit level for the BSP, i.e., Comparing the three options, we find that if then
128 i.e., when H type consumers value their requested content more than the L type consumers beyond a threshold, the BSP benefits from charging the same price to both types and assigning a lower priority to traffic from H type consumers. If on the other hand then ., i.e. when H type consumers and L type consumers have similar valuation for content, the BSP prefers to charge a higher price and in return offer preferential delivery to the data packets from the H t ype con sumers. This leads to P roposition 6 2 Proposition 6 2: ( The ( 1 ) If ; ( 2 ) If ; ( 3 ) If Proof : S ee Appendix B O verall P reference for P ricing S tructure ( C hoice A mong the S ix O ptions NN1, NN2, NN3, NNN1, NNN2, NNN3) In this sub section I address the question of what would be the equilibrium outcome if the BSP is given all six user discrimination options. Proposition 6 3 summarizes the comparison result of all six options.
129 Proposition 6 3: ( The There are two potential preferred pricin g structures for the BSP: NN3 or NNN1, depending on the parameter values. Proof : see Appendix B I illustrate these results by adopting some real life parameter values. AT&T has recently estimated that their top 5% of users (in terms of usage) account for about 40% of the total traffic, i.e., and (Tweney 2008) Using these parameter values, Figure 6 1 depicts in the space The BSP prefers an NN3 outcome (two part tariff with equal priority) in the shaded area and it prefers an NNN1 outcome (uniform fixed fee with low priority for heavy users) in the un shaded area. The area marked by the bold lines represents the feasible parameter space The different intercepts etc. on the two axes (the precise values of these intercepts have been defined in Appendix B) and the straight lines emanating from them represent the different regions (marked by numbers 1 through 7) in the parameter space within which we have to consider the optimal regime choice for the BSP region 1, NN3_2 for region 2, NN3_1 for region 3, NNN1_1 for region 4, NNN1_2 for region 5, NN3 _1 for region 6, and NN3_1 for region 7. A different traffic pattern would change the slope of the line that has the intercept of but would not materially change the nature of the outcome there would still be some regions whe re the BSP would opt for the NN3 outcome and the rest of the feasible region where the BSP would opt for the NNN1 outcome.
130 As outlined before, the choice of the social planner with regards to the pricin g/prioritization regime might be at odds with that of the BSP, since social welfare is the the broadband services are effectively internal transfers as far as t he calculation of the social welfare is concerned, the only measurable effect of the consumers on the social welfare comes from their valuation and the disutility that they attribute towards the congestion. The S ocial P P reference for P ricing S truc ture U nder N et N eutrality (NN1, NN2, NN3) In this sub section, I under net neutrality by comparing the social welfare levels when the BSP adopts the three pricing structures. P ropositi on 6 4 summarizes the analysis. Proposition 6 4: ( The neutrality) When net neutrality is in place, social welfare is the same for one level fixed fee, two level fixed fee, and two part tariff, i.e., This is expected, since the effect of pricing is internalized, and there are no other effects to consider, as traffic prioritization is not allowed under net neutrality. The S ocial P P reference for P ricing S tructure U nder N o N et N eutrality (NNN1, NNN2, NNN3) In this subsection, I under no net neutrality by comparing the social welfare levels when the BSP adopt s the three pricing structures. Proposition s 6 5 and 6 6 summarize the results.
131 Proposition 6 5: ( The neutrality) Without net neutrality, the social planner always prefers the BSP charging a uniform fixed fee while downgrading heavy users, i .e., Proof : See Appendix B The S ocial P O verall P reference for P ricing S tructure (NN1, NN2, NN3, NNN1, NNN2, NNN3) Proposition 6 6: ( The P roof : S ee A ppendix B Differe nces B ocial P P references Based on Proposition s from 6 1 to 6 6, we can see the BSP has incentive to deviate its pricing choice from the social optimum. I summarize the differences in Propositi on 6 7. Proposition 6 7: ( The the social optimum) preference under two scenarios: (1) and ; and (2) and Proof : S ee Appendix B
132 Figure 6 1. The BSP and how it differs from the social N otes: ( 1 ) The area marked by the bold lines represents the feasible parame ter space ( and and where and are defined in the appendix). ( 2 NN3_2 for region 2, NN3_1 for region 3, NNN1_1 for region 4, NNN1_2 for region 5, NN3_1 for region 6, and NN3_1 for region 7. ( 3 ) The shaded areas correspond to preference.
133 T able 6 1. s ix user discrimination o ptions Options Results Net Neutrality NN1 NN2 NN3 Case NN3_1: If Case NN3_2: If
134 Table 6 1. Continued Options Results No Net Neutrality NNN1 Case NNN1_1: If Case NNN1_2: If NNN2 NNN3
135 CHAPTER 7 CONCLUSIONS In this chapter, I conclude my dissertation by summarizing major findings in previous chapters and discussing pote ntial directions for future research Summary of Major Findings Chapter 3 in this dissertation aims to answer two fundamental issues of content provider discrimination: (1) the winn ers and lose rs of abandoning net neutrality; and (2) incentive to expand their capacity without net neutrality I find that if the principle of net neutrality is abolished the BSP definitely stands to gain from the arrangement, as a result of extracting the preferential delivery charge from the content providers. The c ontent providers are thus left worse off, mirroring the stances of the two sides in the debate. Depending on the parameter values in the framework, consumer surplus either does not change or is higher in the short run and in the latter case, while a major ity of consumers are better off, a minority is left worse off with larger wait times to access their preferred content. S ocial welfare in the short run increases when compared to the baseline net neutrality case when one content provider pays for preferent ial treatment, but remains unchanged when both content providers pay. The crucial parameter that determines the nature of the equilibrium is the relative magnitude of the revenue generation capabiliti es of the two content providers: if they differ signific antly, the consumers of in terms of revenue generation) content provider, who are a minority, are left worse off. T he incentive for the broadband service provider to expand capacity under net neutrality is mostly higher than the incentive to expand when the principle of net neutrality is abolished The exception to this outcome occurs when the BSP consumers and only one content provider has the incentive to pay the priority delivery fees. Similarl y, for most of the parameter space, the BSP
136 neutrality (NN) is higher than that under the no net neutrality (NNN) regime. In fact, t he experience in broadband markets around the world indicates that there might be some o ther forces in play that account for the infrastructure capacity expansions in other countries In Japan, for example, fierce competition among broadband service providers has led to the introduction of download bandwidth speeds in excess of 100 Mbps as fa r back as in 2004 (Yang et al. 2004) with prices for the consumers significantly lower than that in the United States (Turner 2005) A fina l finding that should be of interest to policymakers is that under net neutrality, the BSP invests in broadband infrastructure to reach the socially optimal level, but when there is no net neutrality, the BSP either under or over invests in infrastructure In Chapter 4 I relax the full market coverage assumption and thereby simultaneously analyze the effect of abandoning net neutrality principles on the competition between content provi access pricing for consumers In t he process, I design an algorithm that makes an important methodological contribution towards solving a problem that eluded analytical tractability but still had certain structural properties that enabled its deconstruction. From a theoretical perspective, I derive several important results. I find that if net neutrality regulation is not enforced, the BSP will always prefer to charge content providers for preferential delivery of their packets. Further, given the assumption of O utcome 3 ( G paying) is always preferable to the BSP than Outcome 2 ( Y paying) (the resul t would have been reversed if one had assumed ). Finally, I access is above a certain thres hold, the market will be fully covered. More crucially, from a practical perspective, there are several impo rtant implications of this chapter The salubrious effects of NNN include higher market coverage in many cases, and an increased consumer surplus as a whole with certain sets of parameter values. Consumers subscribing to the paying
137 content provider often gain from the arrangement, but within a certain range of parameter values, consumers from the non paying content provider are actually worse off. Wit h content providers effectively subsidizing the consumers for accessing the Internet, access prices are never higher (and are often lower) with NNN. In fact, when consumers do not value their Internet access highly (and with some other conditions holding t rue), the BSP might even provide the Internet service for free. On the flip side, I rivals can effec tively be shut out of the market, as the BSP might find it profitable to institute a priority delivery pricing plan that is too high for them to pay, and in the presence of sufficient congestion in the network, subscribers might abandon that content provid er altogether. Chapter 5 analyzes the issue of vertical integration if a BSP integrated with that of a content provider, will the vertically integrated firm have enough market power to negatively affect the social welfare? I find that the answer crucially depends on whether the BSP is integrated with the more or less effective content provider (effectiveness in this case being defined as the ability to generate advertising revenue from the consumer base). If the vertically integrat ed firm is relatively less effective in generating revenue from its content, I find that for a range of parameter values, the social welfare decreases as compared to a situation when there is no vertical integration. On the other hand, social welfare can a ctually increase if the BSP is vertically integrated with the more effective content provider. But in either case, the competing content provider is often left worse off (and is never better off). In order to decide on an effective policy that regulates ve rtical integration in an online environment, policymakers therefore will need to balance the priorities of increasing social welfare to that of ensuring ef fective competition. Allowing a BSP to form a strategic relationship with the more effective content
138 provider might increase social welfare in the short run. However, this decision might have the less desirable outcome of lowering the level of competition in that particular market segment in the long run, which in turn might lead to less desirable result for the society as a whole. This current research examines the various effects of vertical integration in the presence and absence of net neutrality on social welfare in the short run, future research can look at the effects of vertical integration in the long run where the service capacity becomes a decision variable for the BSP and explore whether the competing content providers are driven out of the market. Other possible areas of exploration include extending the analyses to the duopoly case of two com peting BSP s, relaxing the constraint of market coverage, or allowing for different types of online content from different providers. I n contrast to the extant literature that has looked at the net neutrality problem from the perspective, i n Chapte r 6 I perspective, and examine the economic impact of net neutrality on the BSP and the society as a whole, if the former is not allowed to either prioritize or degrade the delivery of content to one class of users I c onsider scenarios with and without net neutrality and analyze the problem of the monopoly broadband service provider trying out different pricing and prioritization strategies. I find that the impact of net neutrality depends on both the characteristics of the Internet data consumption I find that with net neutrality in place, the BSP would prefer to charge a two part tariff for Internet access, but without net neutrality, a BSP may choose to charge a uniform price a nd degrade heavy users or charge a higher price to high type valuations for content and their usage patterns. Interestingly, I find that w ithout net neutrality in place, degrading the experience of the heavy users increases social welfare a practice that was
139 recently banned by the FCC The joint impact of both pricing and net neutrality under the framework of the proposed model has potentially very important po licy implications. I identify s deviate from the social optimum. enforce net neutrality, there will still be scenarios under which the BSP would opt for a net neutrality solution Future Research S ome immediate areas of future research include consideration of a more sophisticated revenue model for content providers. It is important to note that in orde r to make a meaningful comparison between the two regimes, the content providers must follow the same revenue model under both NN and NNN, and the issue therefore is to choose a revenue model that accurately captures the incentives of the content providers T hanks to content providers like Google or Yahoo!, the overwhelmingly popular revenue model for content providers in the online world is the advertisement assisted model. In this framework, consumers get full access to all the content from the online pro viders. The reason for the popularity of this model is that free content brings about a large number of visitors, who in turn generate revenue by clicking on the advertisements. For several years, many content pro viders aimed for a hybrid model, a prominen t example of which being the online edition of the New York Times whereby some content, which was advertisement assisted, was available for free, but other, ostensibly higher quality content, was made available only to paying subscribers. The New York Tim es abandoned this revenue model in late 2007, after discovering t hat the free, advertisement supported content that was viewed by a large number of users was more profitable than a limited audience of paying subscribers. In fact, Sydell (2007, audio broadcast)
140 subscription model, the online version of the Wall Street Journal, is also expected to make most of its content available for free before the end of 2008 (Anderson 2008) While it has been argued that in the future a significant amount of content would have to be paid for, that right now that remains a conjecture, with a growing body of evidence that the advertisement generated model is becoming more popular with time (Sydell 2007) even with many content providers that would have been expected to follow a subscriber revenue model One example is Qtra x, a new business offering free and legal music downloads with 25 million songs in its inventory Thus, under the current state of affairs, where the hybrid quality differentiated model has been all but abandoned in favor of the purely advertisement assist ed model, advertisement supported free content from the content providers seems to mirror the ground realities. In fact, the prominent technology journalist and author Chris Anderson has argued that this new revenue model based y (Anderson 2008) pointing out that today online content generates revenue from banner advertis ements, affiliate revenues, rental of subscription lists, sale of aggregate information, licensing, live events, listing, paid inclusion, cost per install, getting users to create content for free, streaming audio and video advertising, and API fees, to na me a few (Wilson 2008) Anderson (2008) argues that this is possible today as the Web has made it possible to monetize tw o scarcities that are valuable, reputation and attention, and coupled with the fact that the marginal cost of the content is almost negligible, it has been possible for an increasing number of companies to generate more revenue from the free content (e.g. a free album by Radiohead) than they coul d have by charging for that content on traditional media (charging for that album on a compact disc). In this dissertation I do not consider the capacity (bandwidth) allocation issue. One interesting extension would be to study whether the BSP will find it optimal to partition the
141 capacity and whether such capacity partitioning will change the BSP expanding the infrastructure capacity. A major reason for the BSP to invest less under NNN than under NN is that capacity expansio n reduces the attractiveness of priority delivery of packets for for the priority charge under NNN, which in turn may make BSP desirab le under NNN Similarly, it is of interest to examine the implication of net neutrality on the ability of a content provider to provide premium services (e.g., real time video or remote medical supervision) that require dedicated bandwidth.
142 APPENDIX A LIST OF NOTATIONS : T he marginal consumer indifferent betw een content providers Y and G when the market is fully covered : A n arbitrary consumer on : C : C : Unit price per packet for data packet transmission : C : Poisson arrival rate of content requested from each consumer in packets per unit of time : F it cost for an end consumer away from the ideal content : T he gross value function of r etrie ving content for each consumer when consumers are considered homogeneous : arrive from the websites to the unit cost of delay per uni t of time : T he expected time in the que uing system : C in the short run problem : C : A uniform fixed fee per unit of time charged by the broadband provider to the end consumers : C apacity of the BSP in packets per unit of time : The discount rate used in the long run problem : B in the short run problem : B s profit run problem : Content providers
143 in the long run problem : Consumer surplus : Social welfare : M arginal consumers for content providers Y and G respectively when the full market coverage assumption is relaxed. So the corresponding market shares for content providers Y and G are and respectively. : R evenue rat e per data packet request for the vertically integrated BSP : The vertically integrated BSP : R evenue rat e per data packet request for the independent content provider C in the vertical integration model : C : Content : Percentage of H type consumers : Rate of content requested from H type and L type consumers in packets per unit of time : The gross value of retrieving content for H type and L type consumers respectively : Fixed fees charged to H type and L type consumers respectively : type and L type consumers respectively : The utility function for H type and L type consumers respectively
144 APPENDIX B PROOFS OF PROPOSITIO NS AND OTHER ANALYTI CAL RESULTS Proof of Since is between 0 and 1 and Equation ( 3 3 ) can be rearranged as Define Then solves Since is a convex quadratic function. We also know that and Therefore Proof of Proposition 3 1 (T he R esults of the S hort R un P roblem ) Before we proceed to prove the results for the short run problem, we prove the following important intermediate results which will be repeated ly used in later proofs. It follows from Eq uation ( 3 4 ) that (B 1) (B 2)
145 Subs tituting (B 2) into (B 1) we get (B 3) From F ormula (B 2) and (B 3) we get (B 4) Consumer Surplus under NN Consumer surplus consists of two parts:
146 Social Welfare under NN Internet Access Fee under NNN Case A Substituting (B 4) into the above, we get Therefore
147 BSP under NNN Case A Substituting (B 4) into the above, we get Since we know under NNN Case A Th erefore Consumer Surplus under NNN Case A Since
148 Therefore i.e., t he consumer surplus is increased in NNN Case A compared to NN. Consumer surplus consists of two parts: since since
149 Social Welfare under NNN Case A since Equation (B 4) Therefore i.e., social welfare is increased in NNN Case A compared to NN. Internet Access Fee under NNN Case B
150 BSP under NNN Cas e B Consumer Surplus under NNN Case B Consumer surplus consists of two parts: Therefore the consumer surplus remains unchanged.
151 Social Welfare under NNN Case B Therefore social welfare remains unchanged. Proof of a N ecessary and S ufficient C ondition for the E xistence of a U niqu e It follows from Equation ( 3 11 ) that (B 5) We consider the function We know and achi eves the minimum at with
152 There is a unique solution on if and only if which is equival ent to (B 6) Condition (B 6) is necessary and sufficient for the existence of a unique This condition implies meaning the consumer located at the two ends of the market is lo yal to their corresponding content providers. Under C ondition (B 6) we know Proof of Using implicit differentiation, can be derived from Equation (B 5) by solving
153 Since we know the sign of depends on Since we know from a necessary and sufficient condition for the existence of a unique Proof of Proposition 3 2 ( The B I ncentive to E xpand C apacity) In order to com pare the BSP
154 Substituting and into the above gives The refore Recall that when Outcome 3 is the equilibrium and when Outcome 4 is the equilibrium. We conclude that when or t he BSP has more incentive to expand capacity under NN; when the BSP has more incentive to expand capacity under NNN.
155 Proof of Proposition 3 3 ( The O ptimal C apacity C hoice) Recall that the optimal capacity level for the BSP is determined by first order conditions for NN, for NNN Case A, and for NNN Case B describe d in Equations ( 3 16 ) ( 3 17 ) and ( 3 18 ) respectively. From the proof of Proposition 3 1, when or the BSP has more incentive to expand capacity under NN, i.e., Let and be the optimal capacity levels for N N and NNN respectively. Then and Since Second order condition gives and So and decrease in Therefore Similarly, when the BSP has more incentive to expand capacity under NNN, i.e., Let and be the optimal capacity levels for
156 NN and NNN respectively. Then and Since Second order condition gives and So and decrease in Therefore Proof of Proposition 3 4 ( Whether t he B O ptimal C apacity C hoice is S ocially O ptimal?) The long run social welfare under net neutrality (NN) is The socially optimal capacity is derived by the first order condition Since the BSP that of the long run social welfare. Namely, BSP always invests at the socially optimal level under net neutrality. The long run social welfare under Case A in th e absence of net neutrality (NNN_A) is which can be rewritten as
157 After some algebra, one has We find that when That is, the BSP under invests in capacity If, however, t hen resulting in the BSP over investing in capacity. For the Case B in the absence of net neutrality (NNN_B), the long run social welfare is Since the BSP always under invests under Case B of no net neutrality Proof of Proposition 4 1 (Market Coverage U nder Net Neutrality) and in Formulation ( 4 1 ) can be simplified to and Next we prove that and in Formulation ( 4 1 ) are Proof by co ntradiction: Suppose is an optimal solution. There are three cases. Case 1: and Then substitute by would be a feasible s olution which y ields a higher objective value.
158 Case 2: and Then So Then substitute by and by where such that and would be a feasible solution. Since implies So and the objective value is higher which is a contradiction. Case 3: and Then So Similar to Case 2, we get a contr adiction. Therefore the first two constraints are binding. Since the first two constraints are binding, we get Then the problem can be simplified as: s.t. and
159 Notice that is the total market. T he participation constraint is always binding. So Then the problem becomes: s.t. FOC: Consider function Take the first derivative, we know So decreases in and therefore there is at most one solution to (the FOC condition) in Now check the boundary points: When When Define
160 When there exists an unique where solves the FOC condition When in which means increases in and therefore achieves the highest value on Proof of Lemma 4 1 (Feasibility Condition for Outcome 2) Proof of necessity: From c we get From we get Therefore if Outcome 2 is feasible then Proof of sufficiency: If then there exist a satisfying constraints both and so is a feasible We can easily show is always feasible. Therefore is sufficient for Outcome 2 to be feasible. Proof of Lemma 4 2 (Optimization Condition for Outcome 2) Proof by contradiction. Suppose when Outcome 2 is feasible, Then substitute by would be a feasible solution which yields a higher objective value which is a contradiction.
161 Proof Proposition 4 2 (Outcome 3 D ominates Outcome 2) By substituting and into this F ormulation ( 4 3 ) we can see Formulation ( 4 3 ) has a greater feasible region than Formulation ( 4 2 ) Therefore Outcome 3 always yields at least as high a pro fit for the BSP as Outcome 2. Proof of Lemma 4 4 (Optimization Condition for Outcome 3 ) P roof by contradiction. Suppose when Outcome 3 is feasible, constraint (12) is not binding, i.e., Then substituting by would be a feasible solution which y ields a higher objective value. Proof of Proposition 4 3 (Outcome 4 D ominates Outcome 1) Compare the formulations of Outcome 1 and Outcome 4, we can see that Outcome 1 is equivalent to Ou tcome 4 plus the constraint of Therefore Outcome 4 always yields a profit for the BSP that is at least as high as that it gets under Outcome 1. Proof of Proposition 5 1 (Equilibrium of the G ame with a V ertically I ntegrated BSP) Case A: Outcome 1 and 3 are not equilibria. Considering Outcome 1 first, in the F ormulation ( 5 7 ) the incentive compatibility constraint for BS P is i.e., Since this constraint can never be satisfied. Therefore Outcome 1 is not an equilibrium.
162 Now discussing Outcome 3, in the F ormulation ( 5 11 ) and implies that and respectively. But under Case A there is no feasible which satisfies both constraints. Hence, we only consider Outcome 2 and 4, and note that BSP Outcome 4 is since under C ase A
163 Case B: Similar to Case A, Outcome 1 is not equilibri um Hence we need only to consider Outcomes 2, 3 and 4, and note that now BSP Outcome 4 is since under Case B since Equation (B 4)
164 since under Cas e B, Case C: Firstly, I divide C ase C into two subcases, Case C1: and Case C2: Under Case C1, similar to Case A, Outcom e 1 is not equilibri um In F ormulation ( 5 13 ) we have and which impl y that the BSP s profit is Now, consider Outcomes 2, 3 and 4
165 since under C ase C 1 since under Case C1,
166 Under C ase C 2 similar to Ca se A, Outcome 1 is not equilibri um In F ormulation ( 5 13 ) we have and which imply that the BSP s profit is Now, consider Outcomes 2, 3 and 4
16 7 since under Case C2, Proof of the Results in Table 5 1 Case A: O utcome 2 is the equilibri um I n the F ormulation ( 5 9 ) the optimal solution is Hence, the BSP s profit is C ontent provider C s profit is Consumer Surplus under Case A since
168 Social Welfare under C ase A Case B: Outcome 3 is the equilibri um I n the F ormulation ( 5 11 ) the optimal solution is and Hence, the BSP s profit is C ontent provider C s profit is Consumer Surplus u nder Case B
169 since Social Welfare under Case B Case C1 : Outcome 4 ( Case C1) is the equilibri um I n the F ormulation ( 5 13 ) we have an d Hence, the BSP s profit is C ontent provider C s profit is Consumer Surplus under Case C1
170 Since Social Welfare under Case C1 since Case C2, Outcome 4 ( Cas e C2) is the equilibri um I n the F ormulation ( 5 13 ) we have and Hence, the BSP s profit is
171 C ontent provider C s profit is Consumer Surplus under Case C2 is same as Case C1. Social Welfare under Case C2 Proof of Proposition 5 2 (The W elfare I mpact of N et N eutrality with a V ertically I ntegrated BSP) Case A, since since
172 since Case B, since under Case B,
173 Case C, Both Case C1 and C2 have the same social welfare and customer surplus which remain unchanged compared to the benchmark case. Proof of Proposition 5 3 (The W elfare E ffect of V ertical I ntegration) Case when
174 since Case when
175 since Proof of Proposition 5 4 (The S ocial P P references for V ertical I ntegratio n C ompared to the BSP C hoice) I n Table 5 2,
176 Region 1: Region 2: As we already know we only need to compare to
177 Region 3: As we already know we only need to compare to Region 4:
178 Now we move on to compare to As shown as Figure 5 4 b note that the intersection point of and is which means that i n region 4 if ; otherwise, Now show that I f then
179 since since If then since since
180 Region 5: As we already know we only need to compare to Region 6: As we already know when we only need to compare to Region 7:
181 Solution of Formulation ( 6 7 ) Option NN3 In order to simplify the proofs in Chapter 6 we introduce the following notations. Define
182 We know that since Next we show as below. The intersection between and is The intersection between and is since
183 Now I move on to solve Formulation ( 6 7 ) Option NN3 In Formulation ( 6 7 ) f rom we get From we get We consider two cases: Case 1: Then So constraint is binding, i.e., Substituting into the objective function gives The optimal solution is and Case 2: Then So constraint is
184 binding, i.e., Substituting into t he objective function gives Case 21 : The optimal solution is and Case 22: The optimal so lution is (Since .) The above cases can be summarized as: Case NN3_1: If i.e.,
185 The corresponding consumer surplus is Therefore the social welfare is Case NN3_2: If i.e., The corresponding consumer surplus is
186 Therefore the social welfare i s Proof of Proposition 6 1 ( The P referred P ricing S tructur e U nder N et N eutrality) Now we move on proving Proposition 6 neutrality, we know Then we compare and to If i.e., Since If i.e., since
187 Proof of Proposition 6 2 ( The P referred P ricing S tructure U nder N o N et N eutrality) s three options under no net neutrality, we know Then we compare and to If i.e., then So If i.e., then
188 Since we get if ; If Summarizing the results, we have: if ; if ; if Proof of Proposition 6 3 ( The O verall P referred P ricing S tructure) From the results of Proposition 6 1 and Proposition 6 2, we know : under net neutrality, If If Under no net neutrality, If then ; If ; If Case 1 : If and then
189 If then If and then Case 2 : If an d then If and then If and then Case 3 : If and then
190 So Case 4 : If and then If and then Notice is also where the line intersects the line If and then
191 Case 5 : If and then If i.e., then If i.e., then Case 6 : If and then So
192 Proof of Proposition 6 5 ( The Social P P referred P ricing S tructure U nder N o N et N eutrality) Comparing social welfare under NNN1, NNN2, a nd NNN3 gives Since Therefore Proof of Proposition 6 6 ( The Social P O verall P referred P ricing S tructure) Recall and
193 Since Proof of Proposition 6 7 ( The D eviation from the S ocial O ptimum) From Proposition 6 6, we know that the social planner always prefers NNN1. From Proposition 6 3, we know that under the following two scenarios: (1) and ; (2) and the BSP prefers NN3 and therefore the
194 LIST OF REFERENCES Al Chalabi, M. 2008. "Broadband Usage Based Pricing and Caps Analysis," GigaOm. Anderson, C. 2008. "Free! Why $0.00 is the Future of Business," in: Wired pp. 140 194. Armstrong, M. 2006. "Competition in two sided markets," RAND Journal of Economics (37 :3), pp 668 691. Armstrong, M., and Wright, J. 2007. "Two sided Markets, Competitive Bottlenecks and Exclusive Contracts," Economic Theory (32:2), pp 353 380. Bandyopadhyay, S., and Cheng, H.K. 2006. "Liquid Pricing For Digital Infrastructure Services," In ternational Journal of Electronic Commerce (10:4), pp 47 72. Berners Lee, T. 2006. "Neutrality on the Net." Brander, J.A., and Spencer, B.J. 1983. "Strategic Commitment with R&D: The Symmetric Case," The Bell Journal of Economics (14:1), pp 225 235. Busine ssWire. 2006. "AT&T and Yahoo! Expand Relationship to Provide Consumers Internet Based Voice Communications Services," in: BusinessWire Carr, N. 2008. The Big Switch: Rewiring the World, from Edison to Google W. W. Norton. Cauley, L. 2007. "U.S. Net Acce ss Not All That Speedy," in: USA Today Coalition, H.T.B. 2002. "Ex Parte Submission in CS Docket No. 02 52," FCC (ed.). Dunbar, J. 2006. "'Net Neutrality' would Get Warm Reception in Democratic Congress," in: Milford Daily News p. D2. Dyson, E. 1994. "In tellectual Property on the Net," Edventure Holdings Inc., New York, NY, p. 32. EC. 2006. "Internet: Commission Seeks Global Partnership on Internet Governance, Freedom of Expression and the Combat Against Cyber Repression." European Commission. Economides, N. 2008. "'Net Neutrality,' Non Discrimination and Digital Distribution of Content through the Internet," I/S: A Journal of Law and Policy for the Information Society (4:2), pp 209 233. Economides, N., and Tag, J. 2007. "Net Neutrality on the Internet: A Two sided Market Analysis," in: NET Institute Working Paper #07 45 New York, NY: Stern School of Business, New York University. Economist. 2003. "Between a Rock and a Hard Place," in: The Economist p. Special Supplement.
195 FTC. 2007. "Broadband Connectivit y Competition Policy," F.T. Commission (ed.). Federal Trade Commission, p. 165. Hahn, R., and Wallsten, S. 2006. "The Economics of Net Neutrality," The Berkeley Economic Press Economists' Voice (3:6), June, pp 1 7. Hausman, J.A., Sidak, J.G., and Singer, H .J. 2001. "Residential Demand for Broadband Telecommunications and Consumer Access to Unaffiliated Internet Content Providers," Yale Journal on Regulation (18:1), pp 129 173. Helm, B. 2006. "Tech Giants' Internet Battles," in: BusinessWeek Hermalin, B.E., and Katz, M.L. 2007. "The Economics of Product Line Restrictions with an Application to the Network Neutrality Debate," Information Economics and Policy (19), pp 215 248. Hotelling, H. 1929. "Stability in Competition," The Economic Journal (39:153), March pp 41 57. ITU. 2006. "World Telecommunication/ICT Indicators Database 2006," International Telecommunications Union, Geneva, Switzerland. Jamison, M.A., and Hauge, J.A. 2007. "Getting What You Pay for: Analyzing the Net Neutrality Debate." Knowledge@Whar ton. 2008. "Betting on Betas: How Internet Entrepreneurs Are Creating New Paths to Online Revenue," in: Knowledge@Wharton Krim, J. 2005. "Executive Wants to Charge for Web Speed," in: Washington Post Washington D.C.: p. D05. Krugman, P. 2008. "Bits, Band s and Books," in: The New York Times Lessig, L. 2001. The Future of Ideas: The Fate of the Commons in a Connected World (1 ed.). Random House. Mandy, D.M. 2000. "Killing the Goose That May Have Laid the Golden Egg: Only the Data Know Whether Sabotage Pay s," Journal of Regulatory Economics (17:2), pp 157 172. McCullagh, D. 2007. "Comcast really does block BitTorrent traffic after all." Retrieved May 31, 2009, from http://news.cnet.com/8301 13578_3 9800629 38.html McCullagh, D., and Broache, A. 2006. "Republicans Defeat Net Neutrality Proposal," in: CNet.com McKinseyQuarterly. 2009. "Hal Varian on how the Web challenges managers," in: McKinsey Quarterly
196 Mendelson, H. 1985. "Pricing Comput er Services: Queueing Effects," Communications of the ACM (28:3), pp 312 321. Musacchio, J., Schwartz, G., and Walrand, J. 2008. "A Two Sided Market Analysis of Provider Investment Incentives With an Application to the Net Neutrality Issue," Review of Netw ork Economics (8:1), pp 22 39. National Cable & Telecommunications Association. 2003. "Ex Parte Submission in CS Docket No. 02 52," S. Ms. Marlene H. Dortch, FCC (ed.). Washington, D.C. Owen, B.M., and Rosston, G.L. 2003. "Local Broadband Access: Primum No n Nocere or Primum Processi? A Property Rights Approach." John M. Olin Program in Law and Economics. Representatives. 2005. "Staff Discussion Draft of Legislation to Create a Statutory Framework for Internet Protocol and Broadband Services: Hearings," in: Subcomm. on Subcommittee on Telecommunications and the Internet of the the Committee on Energy and Commerce, House of Representatives, 109th Cong. Washington, D.C. Ricadela, A. 2007. "U.S. Broadband Access Slips Further," in: BusinessWeek Roller, L. H., a nd Tombak, M.M. 1990. "Strategic Choice of Flexible Production Technologies and Welfare Implications," The Journal of Industrial Economics (38:4), pp 417 431. Senate. 2006. "Net Neutrality: Full Committee Hearing," in: U.S. Senate Committee on Commerce, Sc ience, and Transportation., 109th Cong. Washington, D.C. Shapiro, C., and Varian, H.R. 1998. "Versioning: The Smart Way to Sell Information," in: Harvard Business Review pp. 106 114. Shields, M. 2007. "Yahoo Pacts With Comcast.net to Sell Ads," in: Mediaw eek Speta, J.B. 2000a. "Handicapping the Race for the Last Mile?: A Critique of Open Access Rules for Broadband Platforms," Yale Journal on Regulation (17:1), pp 39 91. Speta, J.B. 2000b. "The Vertical Dimension of Cable Open Access," University of Colora do Law Review (71:4), pp 975 1010. Sydell, L. 2006. "Internet Debate Preserving User Parity," in: All Things Considered USA: NPR. Sydell, L. 2007. "Firms Abandon Online Subscription Plans," in: All Things Considered USA: NPR. Cyberspace?," Cato Institute.
197 Turner, S.D. 2005. "Broadband Reality Check: The FCC ignores America's Digital Divide," Consumers Union; Consumer Federation of America; Free Press. Turner, S.D. 2007. "Give Net Neutrality a Chance," in: BusinessWeek Tweney, D. 2008. "AT&T Embraces BitTorrent, May Consider Usage Based Pricing," in: Wired van Schewick, B. 2007. "Towards an Economic Framework for Network Neutrality Regulation," Journal on Telecomm unications and High Technology Law (5:2), pp 329 392. Waldmeir, P. 2006. "The Net Neutrality Dogfight that is Shaking Up Cyberspace," in: Financial Times New York: p. B12. Wilson, F. 2008. "Make Money Around Free Content." Windhausen, J. 2006. "Good Fence s Make Bad Broadband: A Public Knowledge White Paper," pp. 13 16. WSJ. 2006. "Should the Net be Neutral." New York, NY. Wu, T. 2003. "Network Neutrality, Broadband Discrimination," Journal of Telecommunications and High Technology Law :2), p 141. Wu, T. 200 Nondiscriminatory Access"," in: House Committee on the Judiciary Telecom & Antitrust Task Force Washington, D.C. Wu, T. 2006b. "Net Neutrality FAQ." Wu, T., and Lessig, L. 2003. "Re: Ex Par te Submission in CS Docket No. 02 52," S. Ms. Marlene H. Dortch, FCC (ed.). Washington, D.C. Yang, C., Ihlwan, M., and Tashiro, H. 2004. "Commentary: Behind In Broadband," in: BusinessWeek Yoo, C.S. 2006. "Network Neutrality and the Economics of Congestio n," Georgetown Law Journal (94).
198 BIOGRAPHICAL SKETCH Hong Guo earned her Bachelor of Engineering and Master of Engineering in management information systems from Renmin University of China in 1999 and in 2002, respectively. Before she came to Universit y of Florida for her doctoral study, Hong obtained her Master of Science in business administration at the University of Rochester. She received her Ph.D. in business administration from the University of Florida in August 2009 and will join the University of Notre Dame as an assistant professor in Fall 2009.