Business of Social and Other Networks Presentation - PowerPoint Presentation

Business of Social and Other Networks Presentation Prof. Nicholas EconomidesStern School of Business, New York Universityhttp://www.stern.nyu.edu/networks/Haas School of Business, UC Berkeleyand NET Institute http://www.NETinst.org/mailto:economides@stern.nyu.eduApril 29, 2015Copyright ©

Week 1: Introduction Requirements Midterm examClass participation; interactivityFinal group projectGroups of 3 writing an original paperPreliminary presentation of paper on week 10Subject matterAreas we will cover 2

Overview: Areas we will cover Week 1-2, pp. 16-47 (all approx.)Summary of basic economics of competition in non-network industriesBasic Game Theory 3

Overview: Areas we will cover Week 2, pp. 48-75Nature and types of networksVarious types of networks including virtual networksFeatures Network creation/expansionSuperimposition of various networks 4

Overview: Areas we will cover Week 3, pp. 76-85Internet basicsInfrastructureBasic ProtocolsPricingSearch algorithms and advertising“Organic” search; paid searchWeek 4, pp. 86-102Cybersecurity and privacy 5

Overview: Areas we will coverWeek 4: pp. 103-114Network structures and platformsTwo-sided pricing6

Overview: Areas we will cover Week 5, pp. 115-140Network structures and platformsNetwork propertiesNetwork effectsPlatforms Pricing on networksNature of competition on networks 7

Overview: Areas we will cover Weeks 6-7, pp. 141-164Compatibility and interconnectionDesirability Divergence of private and public incentives Public policy issues 8

Overview: Areas we will cover Week 7, pp. 165-179Bottlenecks and Interconnection Pricing 9

Overview: Areas we will cover Week 8, pp. 180-184Taxi cabs as a two-sided networkElectric cars as a two-sided network 10

Overview: Areas we will coverWeek 9, pp. 185-220Payment Systems (credit and debit cards)Mobile Money11

Overview: Areas we will cover Weeks 10, pp. 221-237Digital BooksWeek 11, pp. 238-279Two-sided Pricing & Network NeutralityNetwork neutrality on the Internet 12

Overview: Areas we will cover Weeks 12 Bottom line, pp. 280-304Additional materialsAntitrust in Networks Industries, pp. 305-329Bank networks formation and systemic risk, pp. 330-33613

References Chris Anderson, The Long Tail, Hyperion, New York, 2006. John Battelle, The Search, Penguin, New York, 2005. Brock, Gerald, The Telecommunications Industry: The Dynamics of Market Structure, Harvard University Press, 1981. Carlton, Dennis, and Jeff Perloff, Modern Industrial Organization, Harper Collins. Crandall, Robert, After the Breakup: US Telecommunications in a More Competitive Era, The Brookings Institute, Washington, D.C., 1991. David Easley and Jon Kleinberg, Networks, Crowds, and Markets: Reasoning About a Highly Connected World, Cambridge University Press, 2010 Laffont, Jean-Jacques, and Jean Tirole, Competition in Telecommunications, MIT Press, 2002. Mitchell, Bridger M., and Ingo Vogelsang, Telecommunications Pricing: Theory and Practice. Cambridge University Press, 1991. Owen, Bruce, and Steve Wildman, Video Economics, Harvard University Press, 1992. Shapiro, Carl, and Hal Varian, Information Rules , Harvard Business School Press, 1999. Specific references in each module below. 14

15 Telecommunications (data, voice) Internet / world wide web Social networks on the Internet (Facebook, Twitter, etc.) Broadcasting (TV, radio) Cable television Financial networks Credit and debit card networks ATMs, bank networks; payment systems; check clearing houses Financial exchanges (equities, bonds, derivatives) Banking networks and systemic risk B2B, B2C exchanges Electricity Railroads Airlines Roads Virtual networks Computer software and hardware Information servers (yellow pages, Google, Yahoo, MSN) Network industries are a large part of the world economy and some are growing very fast

A virtual networkis a collection of compatible goods/components that share a common technical platform. For example, all VHS video players make up a virtual network. Similarly, all computers running Windows can be thought of as a virtual network. Compatible computer software and hardware make up a “network,” and so do computer operating systems and compatible applications. More generally, virtual networks are composed of complementary components, so they also encompass wholesale and retail networks, as well as information networks and servers such as telephone yellow pages, Yahoo, Google, etc. 16

17 Network industries often provide necessities provide infrastructure are key to economic growth Network industries have special features

18 Is competition different in these industries? What are successful strategies for companies in markets and industries with network effects? Is there a special case for or against antitrust or competition policy scrutiny for network industries? What form should intervention take (if any)? Price controls? Subsidies? Structural changes (breakups, divestitures, etc.) Imposition of technical compatibility Regulation

Weeks 1-2: Summary of Basic Competition ConceptsSupply and demandDefinitions Costs Consumer surplus (CS)Producer surplus (PS)Total surplus (TS)For more extensive discussion, see Pindyck and Rubinfeld, Microeconomics 19

Demand, supply, and price determination 20

Cost definitions Total costs: C(q).Variable costs: V(q).Fixed costs: F, constant.Breakdown of total costs C(q) = F + V(q).Average total cost: ATC(q) = C(q)/q.Average variable cost: AVC(q) = V(q)/q.Average fixed cost: AFC(q) = F/q.Breakdown of average costs ATC(q) = F/q + AVC(q).Incremental (marginal) cost: MC(q) = C'(q) = dC/dq = V'(q) = dV/dq. 21

Three cost technologies:(1) Constant Returns to Scale 22

Cost technologies: (2) IncreasingReturns to Scale, F > 0, MC constant 23

Cost technologies: (3) Increasing and Then Decreasing Returns to Scale 24

Consumers’ surplusCS = Total willingness to pay for q units minus consumers’ expenditure = Area under demand up to q units minus consumers’ expenditure = A(q) – E(q) E(q) = pq when all units are sold at the same price q, but companies sell in many other pricing schemesKeep in mind that in network industries, the CS calculation will be different 25

26

Producers’ surplusPS = Revenue up to q units minus variable costs = Revenue minus area under marginal cost curve = pq – V Note that Variable Cost V(q) is the area under Marginal Cost MC(q) 27

28

Total surplus TS(q) = CS(q) +PS(q) = A(q) – E(q) + E(q) – V(q) = A(q) – V(q) = net contribution to society of a market for q units (excluding fixed costs)We judge markets according to contribution to TSTS is maximized at qc.Difference TS(q)-TS(qc) = dead weight loss = loss to society from less or more production than qc.Example of less production than qc: monopolyExample of more production than qc: subsidy 29

30

Total surplus is maximized at qc 31 Keep in mind that this will NOT be true in network industries

Price discrimination: ways to reduce consumer surplus and increase profits Sell different units of the same good to the same buyer at different pricesSell different units of the same good to different buyers at different pricesTying: good A not sold without good BApple was not selling Macs without an Apple monitorMixed bundling: selling collections of goods at lower prices than a-la-carteMicrosoft office bundle vs. a-la-carteLoyalty-requirement contract: Discounts are offered if the buyer commits to buying all or most of his “needs” from same seller 32

Price discrimination towards the same buyer Selling different units of the same product to the same buyer at different pricesElectricity bills: higher marginal price for more units In other cases, last unit sold at a lower price (quantity discount)Collecting a fixed fee with or without a variable feeMobile, fixed telecom service bills33

Price discrimination towards the same buyer Loyalty-requirement contract:Offer a discount on some or all units only if buyer buys 90% of his “needs” from youOr offer a discount on some or all units of good A only if the buyer buys 90% of his needs of goods A & BEssentially forces buyer to buy B from dominant firm in A34

Price discrimination towards the same buyerProblems of requirement contracts What if there is a monopolist in good A but there are rivals in good B that produce superior quality of B?Under the requirement contract, rivals in B are eliminated or marginalized with significant TS loss35

Price discrimination towards the same buyer Problems of requirement contracts A multi-unit buyer has some CS left in product A even under monopolyThe monopolist seller can extract more CS by requiring that the product be only sold with B36

Price discrimination across buyers Example: IBM in the 1930s leased tabulating machines; required buyers to buy IBM cardsCard use proportional to machine useMachine use roughly proportional to value to buyerIBM used tying to quantify the value to individual buyerIBM used tying to effectively lease different units of the tabulating machines to different buyers at different pricesOften tying and bundling are illegal under antitrust law 37

Market organization Perfect competitionMany rivalsMarket price is equal to incremental costNo need to take rivals into accountMonopolySingle sellerMarket price significantly exceeds incremental costNo rivalsOligopolyFew firmsMarket price above cost but below monopolyNeed to take rivals’ actions into accountUse game theory to describe market equilibrium 38

Market concentrationHerfindahl-Hirschman index of concentration HHI = Σin=1 si2 Range from 1 (monopoly) to 0 (perfect competitionAt egalitarian equality, all si = 1/n, HHI = 1/nLawyers multiply Σin=1 si2 by 10,000 so for them the HHI range is [0, 10000] 39

Market classifications by HHI defined by DOJ and FTC http://www.justice.gov/atr/public/guidelines/hhi.htmlHHI between 1,500 and 2,500: moderately concentratedHHI > 2,500: highly concentrated“Mergers that increase HHI by more than 200 in highly concentrated markets are presumed likely to enhance market power”40

Game TheoryApplication to Oligopoly Game in extensive form: N players; decision tree; moves; outcomes; information sets; payoffs Π1, Π2; Game in normal form: strategiesPerfect information, perfect recallComplete information; Harsanyi’s analysis of incomplete information 41

Games in extensive form: sequential incumbent-entrant game 42

Games in extensive form: simultaneous games 43

Non-cooperative gamesNon-cooperative equilibrium (Nash equilibrium, NE): A pair of strategies (s1*, s2*) such that no player can increase his payoff provided that the rival does not change his strategy Best reply: b1(s2) maximizes Π1(s1, s2)Best reply: b2(s1) maximizes Π2(s1, s2)At NE, s1* = b1(s2*), s2* = b2(s1) 44

Prisoners’ Dilemma Player 2talksilentPlayer 1Talk(2, 2) (6, 0) Silent (0, 6) (5, 5) 45 Non-cooperative equilibrium is at (2, 2) even though (5, 5) maximizes total utility Players cannot commit to stay silent Other interpretations: arms race; oligopoly

Prisoners’ Dilemma as an Oligopoly Game Player 2high Qlow QPlayer 1High Q(2, 2) (6, 0) Low Q (0, 6) (5, 5) 46 Non-cooperative equilibrium is at (2, 2) even though (5, 5) maximizes total utility Players would like to commit to low Q, but this is a violation of antitrust law

Simultaneous Incumbent-Entrant Game Pl. 2 (I)high Qlow QPl. 1 (E)Enter(-3, 6) (8, 8) Stay Out (0, 18) (0, 9) 47 There are two non-cooperative equilibria at (0, 18) and (8, 8) Notice that pl. 1 wants (8, 8) while pl. 2 wants (0, 18)

Battle of the SexesCoordination Game Player 2balletboxingPlayer 1Ballet(2, 1) (1, 1) Boxing (0, 0) (1, 2 ) 48 Non-cooperative equilibria at (2, 1), (1,1) and (1, 2) Which one will occur?

Matching PenniesZero sum game Player 2headstailsPlayer 1Heads (1, -1) (-1, 1) Tails (-1, 1) (1, -1 ) 49 There is no non-cooperative equilibrium Could use “mixed strategies” – probability distributions over the “pure strategies” {H, T} John Nash (1951): there exists a non-cooperative equilibrium in mixed strategies for any matrix game

Is cooperation needed?To establish a network link between A and B it needs to be profitable for both A and B 50

Week 2: Networks BasicsEasley and Kleinberg. See http://www.stern.nyu.edu/networks/E&K/networks-book-ch01.pdf to -ch24.pdf Sanjeev Goyal, Connections, Princeton University Press, 2007. Chapter 2 51

Networks are composed of complementary nodes and linksThe crucial defining feature of networks is the complementarity between the various nodes and links A service delivered over a network requires the use of two or more network componentsThus, network components are complementary to each other 52

Example: the Information Superhighway (1994) 53

Network structure 54 Nodes: entities Edges/links: relationships Directed or not Features/colors of nodes Features of links

Human connections networks 55 Co-authors Facebook Blog readers Members in organizations political parties professions How important is a specific node Strong and weak ties Cohesion issues What holds the network together? Activity on networks How does it get affected by the network structure? Is behavior influenced by network nearness? How?

Blogs in the 2008 presidential campaign 56

Spread of an epidemic 57

World trade network example from Easley and Kleinberg, Figure 1.8 58

Network structure:Undirected and directed graphs 59

Network structure: Connected components 60

Network structure:Breadth-first search 61

A Simple Social NetworkWith Three Components 62 C D A E B A and E are connected . B and E are not connected, but there is a path between them. A and D are also not connected and there is no path between them. C is an isolated node A is a node with a high degree of centrality

A Star (Hub & Spoke) Network 63 JFK ATL LAX ORD SEA

A Complete Network 64 JFK ATL LAX ORD SEA

Competing Platforms 65 C 3 P 2 P 1 C 2 C 5 C 4 C 1 M 6 M 5 M 4 M 3 M 2 M 1 The Cs are single homing; that is, they belong to just one network (connect to a single platform). P 1 and P 2 are the platforms ( e.g. , payment-card networks) The Ms are multi-homing; that is, they belong to both networks (connect to more than one platform). There is no interconnection between the networks

Competing Platforms with Interconnection (e.g., telecom) 66 C 3 P 2 P 1 C 2 C 5 C 4 C 1 C 11 C 10 C 9 C 8 C 7 C 6 C 5 and C 9 ’s conversation is on net . C 5 and C 6 ’s conversation is off net . P 1 or P 2 may need to pay the other an interconnection fee

Concepts, Goyal pp. 9-24“Neighbors” of j are those nodes with direct links to j A network is “regular” if every node has the same number of links“Degree” of node j is the number of direct connections of j = number of neighbors of j“Degree distribution” is a n-long vector of the fraction of nodes with degree k (= 1, …, n)Variance of the degree distributionRange of degrees 67

Concepts, Goyal pp. 9-24Geodesic “distance” between i and j is the shortest path between them “Degree centrality” is (degree)/(n-1)“Closeness centrality” is (n-1)/(sum of distances of j from all other nodes) 68

Measures of importance of nodes and links Distance“small world” phenomenonCentralityClustering coefficient (of A) is theProbability that two randomly selected friends of A are friends with each otherMethods of partitioning networks“betweenness”: what happens when a link is removedCan we apply this to bank collapses and systemic risk 69

Degrees of separation? 6? 70

Importance of Links:Neighborhood overlap = (Number of nodes who are neighbors of both A and B)/(Number of nodes who are neighbors of either A or B) = 1/6 applied to graph below: 71

72

Densely-connected, homogeneousparts that are weakly connected to each other 73

Homophily p fraction of all individuals are maleq (=1 - p) fraction of all individuals are femaleConsider a particular edge in this networkIf we independently assign each node the gender male with probability p and the gender female with probability q, then both ends of the edge will be male with probability p^2, and both ends will be female with probability q^2 = (1-p)^2If the first end of the edge is male and the second end is female, or vice versa, then we have a cross-gender edge, so this happens with probability 2pq = 2p(1-p)Test for homophily according to gender as follows:Homophily Test: If the fraction of cross-gender edges is significantly less than 2pq, then there is evidence for homophily. 74

Homophily example 5 of the 18 edges in the graph are cross-gender. p = 2/3, q = 1/3. We should compare the fraction of cross-gender edges to the quantity 2pq = 4/9 = 8/18. With no homophily, expect to see 8 cross-gender edges rather than 5, and so this example shows some evidence of homophily. 75

Week 3: The InternetEconomides, Nicholas (2007), The Economics of the Internet, in The New Palgrave Dictionary of Economics, London: Macmillan. Economides, Nicholas (2006), The Economics of the Internet Backbone, in Handbook of Telecommunications. Amsterdam: Elsevier Publishers. Easley and Kleinberg, ch. 13, 14, 15. 76

Basics on the Internet Creation; History; Arpanet, NSFStructure, infrastructureTransmission; TCP/IP (transmission control protocol, UDP (user datagram protocol)CommercializationPricing: transit and peeringInternet BackboneMerger of MCI with WCOMBlocked merger of MCI with SprintPresent industry structure 77

Originally the Internet was connecting computers of defense contractors and DoD computers “Packet-switched” not “circuit-switched”Information is cut to small “information packets”Based on simple transmission protocolsTCP/IP (transmission control protocol)Guarantees delivery but may have delaysUDP (user datagram protocol)Immediate delivery, but packets may be lostBecause of nuclear attack threat at the timePackets from A to B are sent in many routesLow security because nodes “knew” each other78

DoD gave it to the National Science Foundation (NSF)Later, in the early 1990s, commercialization Explosive growth; 1 bil devices (hosts)Backbone (long distance) networkInternet Service Providers (“ISPs”) connecting to backboneResidentialBusiness customersUpstream ISPs79

80 Explosive growth of the Internet

Pricing by Internet Backbones (IBPs) to Internet Service Providers (ISPs)“transit” contract, delivers the whole Internet through a virtual pipe of a maximum bandwidth “peering” allows free exchange of info packets between two ISPs X and Y or two backbonesPeering is limited to traffic originating or terminating at X or Y81

82

US Lagging in Broadband 83

USA Lags Behind Poorer Countries 84

BB market structure: US Carrier Traffic in Petabytes per month, May 2005  Company TrafficMarket Share Among All Providers   1Q2004 2Q2004 3Q2004 4Q2004 4Q2004 A (AT&T) 37.19 38.66 44.54 52.33 12.58% B 36.48 36.50 41.41 51.31 12.33% C 34.11 35.60 36.75 45.89 11.03% MCI 24.71 25.81 26.86 30.87 7.42% E 18.04 18.89 21.08 25.46 6.12% F 16.33 17.78 17.47 19.33 4.65% G 16.67 15.04 14.93 15.19 3.65% Total traffic top 7 networks 183.53 188.28 203.04 240.38 57.78%   Total traffic all networks 313 313 353 416   85

Week 3: Cybersecutity and PrivacyNicholas Economides et al. (2010), Toward Better Usability, Security, and Privacy of Information Technology, Committee Report, National Academies of Sciences. 86

IssuesAt present, the incentives of both users and companies on usable security and privacy diverge from actions that would maximize social benefit What economic and legal policies can be implemented to change users’ and companies’ incentives so that they are closer to maximizing social benefit? 87

Significant deficit in usable security and privacy Current operating systems of PCs, netbooks, mobile phones, and other devices have significant security deficienciesInterfaces defining security levels are typically very difficult to followUsers are typically unaware of their level privacy (or its lack) in computers 88

The Internet has multiplied security problems of connected devices and highly increased the global impact of local lack of security The Internet was designed for a small number of nodes that knew and trusted each otherPresently we are almost at a billion nodes world-wide with no mutual knowledge and no trustThe Internet has no centralized or Internet Service Provider (ISP)-level securitySecurity issues have an even more severe impact in “cloud computing”Typical users have a very limited understanding of the network capabilities of their computers and the possibilities of abuse in a network setting 89

Perspectives of the IssueThe residential user’s point of view The business user point of viewA search engine’s point of viewThe network’s (societal) point of viewOperating systems (OSs) and computer manufacturers point of viewISP’s point of viewGlobal issues 90

The user’s point of the view Different computer communications, usage, and storage require different levels of securityDoes the user understand how secure or insecure his communications, usage, and storage are?Does the user understand the financial consequences to him and others of lack of security in these? How can the user’s understanding be enhanced?Does the user have sufficient economic incentives (rewards/punishments) to use sufficient security?What is the balance between the user’s desire for privacy and the user’s desire for communication in social networks?Can we improve usability of security so that users who aim for higher security are able to achieve it? How? 91

Private companies have diverse points of view on security and privacy 1. Some businesses (e.g. banks, stock brokers, electronic commerce firms) generally desire higher securityThey have found various (private) solutions attempting to make their transactions more secure2. But advertisers and search engines generally like more disclosure of private information to be able to pitch products closer to a consumer preferences and willingness to payA very secure Internet where users are fully aware of the impact of disclosures of their private information would cut into the profits of these companies 92

Private companies perspective (cont.) 3. OS companies typically grew up in the pre-Internet eraOSs originally relied on third parties to beef up securityOSs did not anticipate the potential for global damage created by multiple local infiltrations in a network setting in the presence of even small security flawsUltimately, companies will act to avoid liabilityHow should we tweak the law to change the incentives of OS and computing devices manufacturers? Bottom line: Given the diverse uses of the Internet and the various functions/roles of firms on the Internet, it is unlikely to have a consensus among companies on security and privacy 93

From the network’s point of view (societal point of view) In general, the value of security is much higher for the network than for an individual userUsers, left on their own, will generally tend to achieve lower security than society desiresLow security at a node can lead to catastrophic network events (such as the collapse of attacked nodes or even parts of the Internet) that are much more damaging to society than to the individual nodeThe lack of security at a node is a “negative externality” to the network 94

In a network setting there is a crucial divergence between private and social incentives on security Presently most users do not have sufficient incentives to secure their computers to prevent network-wide catastrophic eventsCan we create sufficient economic incentives so that users aim for sufficient security? How?How can we improve usability of security so that users who aim for higher security are able to achieve it? 95

What incentives will induce users to more secure computing behavior?Positive monetary incentives (pay people)? Awards and other non-monetary positive incentives?Punishments for not meeting a security benchmark?Impose on insecure nodes liability for damages created using their node?Limit access to the Internet if computer fails security test? 96

From the OSs’ and computer manufacturers’ point of viewHow can we create incentives for computer and operating systems vendors to increase security and maintain it through the useful life of a computer? 97

Should we require OSs to include and automatically update for free security/antivirus/anti-phishing? Should we impose additional liability on operating systems vendors? In the extreme, should we deny computers access to the Internet (except the security checkup and upgrade site) unless they have passed a minimal standard of security?Should we require OSs to disable (as the default) various server functions of new computers, network devices, mobile phones, etc.? 98

The ISPs point of viewHow can we induce ISPs to play a role in limiting or preventing some attacks while adhering to network openness and net neutrality? 99

Global IssuesNo matter how good security becomes within the U.S., security issues will remain because of the global nature of the Internet This underlines the importance of certification of web sites and of measures that improve security in bilateral communications (including web browsing)of requirements on computer and OS manufacturers to increase security and automatically maintain it through the useful life of consumers worldwide 100

Policy ChangesTo strengthen usable security, what legal and economic policy changes are required at the user level?at the computer and OS manufacturer level?at the web site/server level?at the Internet service provider level? 101

Some questionsHow will the society deal with the “negative externality” on the network/society created by the lack of usable security of individual nodes? How can we provide positive and negative, monetary and non-monetary incentives to users to eliminate the negative externality?What role can the OSs play? Design? Post-purchase security maintenance?What role can the search engines play in making people aware of privacy issues?What role can ISPs play? 102

Week 4: Network structures and platforms; two-sided pricing Economides, Nicholas The Economics of Networks, International Journal of Industrial Organization, vol. 16, no. 4, pp. 673-699 (October 1996). Economides, Nicholas, and Charles Himmelberg, (1995), Critical Mass and Network Evolution in Telecommunications, in Gerard Brock (ed.), Telecommunications Policy Research Conference Selected Papers 1995. Shapiro and Varian chapter 7. 103

104 Special features of markets with network effects Increasing returns to scale in consumption (network effects) A market exhibits network effects when the value of a transaction (or subscription) is higher when more units change hands, everything else being equal

105 Special features of markets with network effects: complementarities In a traditional network, network externalities arise because a typical subscriber can reach more subscribers in a larger network

106 Special features of markets with network effects: complementarities Network effects arise because of complementarities When customer A makes a phone call to customer B, he uses both AS and BS Although goods “access to the switch” AS, BS, …, GS have the same industrial classification and traditional economics would classify them as substitutes, they are used as complements 106

One-way and two-way networksNetworks where services AB and BA are distinct are called “two-way” networks Two-way networks include railroad, road, and many telecommunications networks When one of AB or BA is unfeasible, or does not make economic sense, or when there is no sense of direction in the network so that AB and BA are identical, then the network is called a one-way networkIn a typical one-way network, there are two types of components, and composite goods are formed only by combining a component of each type, and customers are often not identified with components but instead demand composite goodsFor example, radio and TV broadcasting and early paging networks are one-way networks 107

A long distance (two-way) network or an ATM (one-way) network 108

Virtual network or platform; components Di complementary with components Sj; also has network effects 109 In a virtual network, externalities arise because larger sales of components of type D induce larger availability of complementary components S 1 , ..., S n , thereby increasing the value of components of type D Examples: CPUs & monitors Users and advertisers in Yellow Pages or Internet search engines Razors and blades Cameras and film

Financial and other exchanges: network effects arise because market liquidity is desirable 110

Platforms 111

112 Technology platforms are the hubs of the value chains in technology industries Examples of technology platforms: Microsoft Windows (PC operating systems) Intel processors (PC hardware) Sony PlayStation (game consoles) A platform forms the framework on which complementary goods (applications) attach (run)

113 Complementarity between a platform and applications A technology platform may be proprietary or open source Platform examples: Windows, Linux, iOS Application examples: MS-Office, Open Office, iApps

114 Firms can make money from either side of a network or from both sides from a server or a client (example: Adobe Acrobat and Adobe Reader) from a caller (typical) or a receiver (800 numbers) of a phone call or from both (cellular in U.S.) Internet backbones collect money from both parties that send and receive traffic (when not in a peering relationship) The availability of prices on both sides of the network allows for complex pricing strategies, and, depending on the dynamics and market shares on the two sides of the market, can be used strategically to enhance and leverage a firm’s strong strategic position on one side of the network Two-sided pricing in markets with network effects

Weeks 5: Network Effects Under (Technical) CompatibilityEconomides, Nicholas The Economics of Networks, International Journal of Industrial Organization, vol. 16, no. 4, pp. 673-699 (October 1996). Economides, Nicholas, and Charles Himmelberg, (1995), Critical Mass and Network Evolution in Telecommunications, in Gerard Brock (ed.), Telecommunications Policy Research Conference Selected Papers 1995. Shapiro and Varian chapter 7. 115

Complementarity and compatibility Links on a network are potentially complementary, but it is compatibility that makes complementarity actualSome network goods and some vertically related goods are immediately combinable because of their inherent properties However, for many complex products, actual complementarity can be achieved only through the adherence to specific technical compatibility standardsThus, many providers of network or vertically-related goods have the option of making their products partially or fully incompatible with components produced by other firms This can be done through the creation of proprietary designs or the outright exclusion or refusal to interconnect with some firmsAs we will see, it is not always in the best interests of a firm to allow full compatibility of its products with those of its competitors 116

Network effectsNatureHow they arise Effects Consequences on competition, market structure and profits 117

Sources of Network Effects (1) In traditional non-network industries, the willingness to pay for the last unit of a good decreases with the number of units sold. This is called the law of demand, and is traditionally considered to hold for almost all goodsHowever, the existence of network effects implies that, as more units are sold, the willingness to pay for the last unit may be higherThis means that for network goods, the fundamental law of demand is violated: for network goods, some portions of the curve demand can slope upwardsFor some portions of the demand curve, as sales expand, people are willing to pay more for the last unit 118

Sources of Network Effects (2)The law of demand is still correct if one disregards the effects of the expansion of sales on complementary goods But, as increased sales of a network good imply an expansion in the sales of complementary goods, the value of the last unit increasesCombining the traditional downward slopping effect with the positive effect due to network expansion can result in a demand curve that has an upward-slopping partThe key reason for the appearance of network effects is the complementarity between network components 119

Sources of Network Effects (3) Depending on the network, the network effect may be direct or indirectWhen customers are identified with components, the network effect is direct Consider for example a typical two-way network, such as the local telephone networkIn this n-nodes 2-way network, there are n(n - 1) potential goods. An additional (n + 1th) customer provides direct network effects to all other customers in the network by adding 2n potential new goods through the provision of a complementary link to the existing linksIn typical one-way networks, the network effect is only indirect When there are m varieties of component A and n varieties of component B (and all A-type goods are compatible with all of B-type), there are mn potential composite goodsAn extra customer yields indirect network effects to other customers, by increasing the demand for components of types A and B 120

Sources of Network Effects (4) Exchange networks (financial networks such as the NYSE and NASDAQ, commodities, futures, and options exchanges as well as business to business “B2B” exchanges) also exhibit indirect network effectsThere are two ways in which these network effects arise:Network effects arise in the act of exchanging assets or goods Network effects may arise in the array of vertically related services that compose a financial transaction.These include the services of a broker, bringing the offer to the floor, matching the offer, etc. The second type of network effects are similar to other vertically-related markets The first way in which network effects arise in financial markets is more importantThe act of exchanging goods or assets brings together a trader who is willing to sell with a trader who is willing to buyThe exchange brings together the two complementary goods, “willingness to sell at price p” (the “offer”) and “willingness to buy at price p” (the “counteroffer”) and creates a composite good, the “exchange transaction” 121

Networks in finance ReferencesEconomides, Network Economics with Application to Finance (1993) Economides and Schwartz, Electronic Call Market Trading (1995)Economides and Schwartz, Equity Trading Practices and Market Structure: Assessing Asset Managers’ Demand for Immediacy (1995)Lange and Economides  A Parimutuel Market Microstructure for Contingent Claims Trading (2000)122

Sources of Network Effects (5)The two original goods were complementary and each had no value without the other one Clearly, the availability of the counteroffer is critical for the exchange to occurPut in terms commonly used in Finance, minimal liquidity is necessary for the transaction to occurFinancial and business-to-business exchanges also exhibit positive size externalities in the sense that the increasing size (or thickness) of an exchange market increases the expected utility of all participantsHigher participation of traders on both sides of the market (drawn from the same distribution) decreases the variance of the expected market price and increases the expected utility of risk-averse tradersHigher liquidity increases traders’ utility 123

Network Effects Under Compatibility and Perfect Competition (1) Let the willingness to pay for the nth unit of the good when ne units are expected to be sold be p(n; ne)n and ne are normalized so that they represent market coverage, ranging from 0 to 1, rather than absolute quantitiesWillingness to pay p(n; ne) is a decreasing function of n because the demand slopes downwardp(n; ne) increases in ne; this captures the network externalities effect, i.e., that the good is more valuable when the expected sales ne are higher 124

Network Effects Under Compatibility and Perfect Competition (2) At a market equilibrium of the simple single-period world, expectations are fulfilled, n = ne, thus defining the fulfilled expectations demand p(n, n)Each willingness-to-pay curve p(n, nie), i = 1, 2, ..., shows the willingness to pay for a varying quantity n, given an expectation of sales ne = nie. At n = nie, expectations are fulfilled and the point belongs to p(n, n) as p(nie, nie)Thus p(n, n) is constructed as a collection of points p(nie, ni e ) 125

When there are strong network effects, demand can slope upwards The “law of demand,” i.e., that higher output can be sold only at lower prices, is violated when there are significant network effects: demand curve can slope upwards 126

Economides and Himmelberg, Critical Mass and Network Evolution in Telecommunications (1995) show that the fulfilled expectations demand is increasing for small n if either one of three conditions hold: (i) the utility of every consumer in a network of zero size is zero; or (ii) there are immediate and large external benefits to network expansion for very small networks; or (iii) there is a significant number of high-willingness-to-pay consumers who are just indifferent on joining a network of approximately zero sizeThe first condition is straightforward and applies directly to all two-way networks, such as the telecommunications and fax networks where the good has no value unless there is another user to connect toThe other two conditions are a bit more subtle, but commonly observed in networks and vertically-related industries 127

The second condition holds for networks where the addition of even few users increases significantly the value of the networkA good example of this is a newsgroup on an obscure subject, where the addition of very few users starts a discussion and increases significantly its value The third condition is most common in software marketsA software application has value to a user even if no one else uses it The addition of an extra user has a network benefit to other users (because they can share files or find trained workers in the specifics of the application), but this benefit is smallHowever, when large numbers of users are added, the network benefit can be very significant 128

Possibility that the demand does not start at zeroFor goods that have a positive stand–alone value k without network effects, the demand starts at (0, k). 129

Critical Mass When the fulfilled expectations demand increases for small n, we say that the network exhibits a positive critical mass under perfect competition.If we imagine a constant marginal cost c decreasing as technology improves, the network will start at a positive and significant size no (corresponding to marginal cost co)For each smaller marginal cost, c < co, there are three network sizes consistent with marginal cost pricing: a zero size network; an unstable network size at the first intersection of the horizontal through c with p(n, n); and the Pareto optimal stable network size at the largest intersection of the horizontal with p(n, n) 130

131 Market penetration of innovations is much faster in network industries than in non-network industries Penetration time Diffusion of an innovation with and without network effects Non-network industry network industry

Multiplicity of Equilibria The multiplicity of equilibria is a direct result of the coordination problem that arises naturally in the typical network externalities model The existence of an upward slopping part of the demand curve and the multiplicity of equilibria even under perfect competition also allows for a network to start with a small size and then expand significantlySuppose, for example, that marginal cost is at c < co and a new invention creates a new product with significant network effects Then, it is possible that the industry starts at the left intersection of the horizontal at c with p(n, n) as expectations are originally low, and later on advances suddenly and quickly to the right intersection of the horizontal at c with p(n, n)Thus, the multiplicity of equilibria in network industries can lead to sudden significant expansions of network size 132

Efficiency (1)In the presence of network externalities, it is evident that perfect competition is inefficient The marginal social benefit of network expansion is larger than the benefit that accrues to a particular firm under perfect competitionPerfect competition (p = MC) will provide a smaller network than is socially optimal, and, for some relatively high marginal costs, perfect competition will not provide the good while it is socially optimal to provide it 133

Gross benefit (area under the demand) is B(n, n) Marginal benefit of network expansion   134

Efficiency (2)Since perfect competition is inefficient, state subsidization of network industries is beneficial to society The Internet is a very successful network that was subsidized by the US government for many yearsThe subsidized Internet was aimed at promoting interaction among military research projects During the period of its subsidization, almost no one imagined that the Internet would become a ubiquitous commercial networkThe foundation of the Internet on publicly and freely available standards has facilitated its expansion and provided a guarantee that no firm can dominate it 135

136 Complex pricing: externalities internalized or not? Often the additional subscriber/user is not rewarded for the benefit that he/she brings to others by subscribing Hence there may be “externalities,” i.e., benefits not fully intermediated by the market In some cases, externalities are fully intermediated through non-linear pricing Example: Cantor Fitzgerald pricing towards Salomon Brothers in secondary U.S. bonds market (before 2001) Typical trader paid $20 per $1 million face value Salomon paid $1 per $1 million face value plus a fixed fee Why? Salomon brought immense liquidity to the secondary market because it controlled 40% of the primary market 136

Mathematical example p(q, ne) = (1 – q)neq: actual units soldne: expected size of salesFulfilled expectations demand: p(n, n) = (1-n)nMarginal benefit of network expansion = n(1 - n) + n(1 - n/2) = n (2 – 3n/2 ) Notice that dB/dn > p.   137

138 The marginal benefit of network expansion is always higher than the willingness to pay of the last participant

If the good has a stand-alone value k > 0p(q, ne) = (1 - q)(k + ne)p(n, ne) = c  n = 1 – c/(k + ne)p(n, n) = (1 - n)(k + n)dB(n, n)/dn = (1-n)(k+n)+n(1-n/2)Perfect comp: p(n, n) = c  n*(c) = [1-k+((1+k)^2-4c))]/2139

Fulfilled expectations 140

141 Adoption S-curves from various network industries

Weeks 6-7: Technical Standards Competition and the Compatibility DecisionPaul David and Shane Greenstein (1990), The Economics of Compatibility StandardsNicholas Economides and Fredrick Flyer, (1998) Compatibility and Market Structure for Network Goods 142

Why is technical standards competition important?Because of network effects Network effects create inequalityCompeting standards that have small market share can be marginalizedWhen a company chooses to be incompatible, it makes a big difference where it stands in the inequality chain of incompatible firms 143

Technical standards competitionVHS vs. Beta Windows vs. Mac vs. LinuxMP3 vs. WMA vs. RealAudioHD DVD vs. Blu-RayiPhone vs. Android vs. Windows 8 144

Technical Compatibility when various links and nodes on the network can be costlessly combined to produce demanded goodsTwo complementary components A and B are compatible when they can be combined to produce a composite good or serviceExample: we say that a VHS-format video player is compatible with a VHS-format tapeTwo substitute components A1 and A2 are compatible when each of them can be combined with a complementary good B to produce a composite good or service Example: two VHS tapes are compatible; two VHS video players are compatibleSimilarly we say that two software products are compatible (more precisely two-way compatible) when they each can read and write files in a common format Compatibility may be one-way when the files of format B1 of software A1 can be read by software A2 , but the files format B2 of software A2 cannot be read by software A1Moreover, compatibility may be only partial in the sense that software A1 is able to read files of format B2 but unable to write files in that format. 145

146 Dichotomy in markets with network effects (1) Full compatibility networks Voice telecommunications (by regulation) Internet data communications (by design) Fax (by design) Cars and gasoline (by market evolution) Tables and chairs (by market evolution)

147 Dichotomy in markets with network effects (2) Incompatible networks Operating Systems for PCs (Windows, Mac OS X, Linux) Game platforms (Xbox, Sony, Nintendo) Digital audio formats (iPod, Windows Media Player “WMA,” MP3, RealAudio) High definition DVDs (HD-DVD, Blu-ray) Video players (Betamax, VHS) Information servers (Google, MSN, Yahoo, yellow pages) Financial and other exchanges

148 Path-dependence is the dependence of a system or network on past decisions of producers and consumers Today’s sales of Windows are path-dependent because they depend on the number of Windows sold earlier (the installed base Windows). The existence of an installed base of consumers favors an incumbent However, competitors with significant product advantages or a better pricing strategy can overcome the advantage of an installed base Example: VHS overcame Beta after six years of higher installed base by Beta Sony’s mistakes in disregarding network externalities and not licensing the Beta format JVC’s widespread cheap licensing of VHS A low-end, low-price VHS player can contribute as much to the network effect as a high-end high-price Beta player

Strategic Choices of Technical Standards and Compatibility In Network Industries Standards Wars (1)A key strategic decision for a firm is the extent to which it will be compatible with other firmsA network good has higher value because of the existence of network effectsDifferent firms conforming to the same technical standard can create a larger network effect while still competing with each other in other dimensions (such as quality and price)The decision to conform to the same technical standard is a strategic oneA firm can choose to be compatible with a rival and thereby create a larger network effect and share it with the rival. A firm could alternatively choose to be incompatible with the rival, but keep all the network effects it creates to itself 149

Standards Wars (2)Which way the decision will go depends on a number of factors: In some network industries, such as telecommunications, interconnection and compatibility at the level of voice and low capacity data transmission is mandated by lawThe decision will depend on the expertise that a firm has on a particular standard (and therefore on the costs that it would incur to conform to it)The choice on compatibility will depend on the relative benefit of keeping all the network effects to itself by choosing incompatibility versus receiving half of the larger network benefits by choosing compatibility.The choice on compatibility depends on the ability of a firm to sustain a dominant position in an ensuing standards war if incompatibility is chosenThe compatibility choice depends on the ability of firms to leverage any monopoly power that they manage to attain in a regime of incompatibility to new markets. 150

Standards Wars (3)Standards may be defined by the government (as in the case of the beginning of the Internet), a world engineering body (as in the case of the FAX), an industry-wide committee, or just sponsored by one or more firms Even when industry-wide committees are available, firms have been known to introduce and sponsor their own standardsIncentives of firms to choose to be compatible with others; coordination game 151

Standards war leading to compatibility (1) Full compatibility at both non-cooperative equilibriaStandard 1 is a non-cooperative equilibrium if a > e, b > d. Similarly, standard 2 is an equilibrium if g > c, h > f. In this game, we will assume that firm i has higher profits when “its” standard i get adopted, a > g, b < h. Profits, in case of disagreement, will depend on the particulars of the industry. One standard assumption that captures many industries is that in case of disagreement profits are lower than those of agreeing on either standard, e, c < g; d, f < b 152

Standards war leading to compatibility (2)There is no guarantee that the highest joint profit standard will be adopted Since consumers surplus does not appear in the matrix, there is no guarantee of total surplus maximization at equilibrium 153

Standards war leading to incompatibility Compatibility with competitors brings higher network externality benefits (“network effect”) and therefore is desirable. At the same time, compatibility makes product X a closer substitute to competing products (“competition effect”), and it is therefore undesirable. In making a choice on compatibility, a firm has to balance these opposing incentives. Firms want to differentiate their products because they want to avoid intense competition.In a network industry, the traditional decisions of output and price take special importance since higher outputInequality in market shares and profitability is a natural consequence of incompatibilityUnder incompatibility, network externalities act as a quality feature that differentiates the products 154

155 Markets with strong network effects where firms can choose to be incompatible are “ winner-takes-most” markets In these markets, there is extreme market shares and profits inequality The market share of the largest firm can be a multiple of the market share of the second largest, the second largest firm’s market share can be a multiple of the market share of the third, and so on Example: 66%, 22%, 7%, 2.%, 1%, … Geometric sequence of market shares implies that, even for small n, the nth firm’s market share is tiny Examples: PC operating systems market; software applications markets Why? A firm with a large market share has more complementary goods and therefore its good is more valuable to consumers Why “winner-takes-most” and not “winner-takes-all”? Because to “take all” requires an undesirable cut in price

“Winner-takes-most” markets When fixed costs are small, a very large number of firms can survive, but there is tremendous inequality in market shares, prices, and profits among themExamples of this market structure are the PC operating systems market and many software applications marketsSetup of Economides and Flyer (1998)All firms produce identical products, except for what value is added to them by network effectsNo firm has any technical advantage in production over any other with respect to any particular platform and no production costs“Pure network goods” where there is no value to the good in the absence of network externalitiesConsumers are differentiated in their willingness to pay for the network good 156

157 With network effects, natural inequality: “winner-takes-most” Markets for incompatible products have inequality Hits in blogs Hits in Internet engines Market share of firms in traditional Yellow pages Size distribution of connections of Internet hosts Because of natural inequality in the market structure of network industries, there should be no presumption that anti-competitive actions are responsible for the creation of market share inequality or very high profitability of a top firm No anti-competitive acts are necessary to create this inequality

Profits inequality and network effects under incompatibility. k = value of the good with no network effects 158

Herfindahl-Hirschman (H) Index for Different Intensities of Marginal Network Externality 1/k and Numbers of Firms S Under Incompatibility ∞ .463No network effect (1/k = 0) implies equal firm shares, si = 1/n and H index:H = Σin=1 si2 = 1/n No network effect 0 .333 .2 .1 0 159

160 Inequality of market shares and prices under incompatibility for pure network goods (k = 0)

Profits, Consumers’ and Total Surplus under incompatibility for pure network goods (k = 0) Ratios of profits of consecutive firms range from 15 to 20 161

Monopoly May Maximize Total Surplus When here are fewer firms in the market there is more coordination and the network effects are largerAs the number of firms decreases, the positive network effects increase more than the dead weight loss, so that total surplus is maximized in a monopoly! Total surplus is highest while consumers’ surplus is lowest in a monopolyThis poses an interesting dilemma for antitrust authoritiesShould they intervene or not? In non-network industries, typically both consumers’ and total surplus are lowest in a monopolyIn this network model, total maximizing consumer’s surplus would imply minimizing total surplus 162

No Anti-Competitive Acts are Necessary to Create Market Inequality In network industries, free entry does not lead to perfect competitionAntitrust and competition law have placed a tremendous amount of hope on the ability of free entry to spur competition, reduce prices, and ultimately eliminate profits In network industries, free entry brings into the industry an infinity of firms but it fails miserably to reduce inequality in market shares, prices and profitsEntry does not eliminate the profits of the high production firmsImposing a “competitive” market structure is likely to be counterproductive 163

Strategic Choice of Compatibility in Duopoly Equilibria in a Two-Firm Industry k 164

165 Intense competition on which firm will create the top platform and reap most of the benefits Example: Schumpeterian races for market dominance among dot-coms in 1999-2000 Very high valuation of dominant vs. other dot-com firms in that period; Wall Street perception Strategic effect: firms advertised very intensely and subsidized consumers to achieve a dominant position Competition for the market takes precedence over competition in the market

Week 7: Bottlenecks and Interconnection Pricing Nicholas Economides, Giuseppe Lopomo and Glenn Woroch, Strategic Commitments and the Principle of Reciprocity in Interconnection Pricing, chapter 5 (pp. 62-99), in Gary Madden (ed.) The Economics of Digital Markets, Edward Elgar (2009).Nicholas Economides, Giuseppe Lopomo and Glenn Woroch, Regulatory Pricing Policies to Neutralize Network Dominance, Industrial and Corporate Change, vol. 5, no. 4, pp. 1013-1028, (1996). 166

167 One-sided bottlenecks The early (1900) AT&T owned links 1 (long distance) and 2 (local), but did not allow independent firms which possessed link 3 to interconnect at B and provide part of the long distance service ABC For over two decades in the beginning of the 20th century, AT&T refused to interconnect independent local telecommunications companies to its long distance network, unless they became part of the Bell System, resulting in 89% market share for AT&T by 1935 from about 50% in 1914 In the 1970s AT&T controlled 89% of local lines and almost all long distance (LD) AT&T faced competition in LD by MCI which did not have local lines AT&T provided low quality interconnection to local lines to MCI

Interconnection Pricing;Vertical Price Squeeze Assume use of link 1 is required by links 2, 3Goods 12 and 13 and substitutes Assume link 1 is monopolized by firm A which also owns link 2, while firm B owns link 3Firm B owns only link 2 and needs to buy use of link 1 from AFirm A sets the price p(12) of end-to-end service 12 and the “interconnection” or “access” price p(1) of link 1 when sold to firm BThe price-to-cost margin of B is p(12) – p(1), both controlled by firm A, and can be made as small as firm A wantsFirm B can be marginalized of driven out of business 168

169 Two-sided bottlenecks Each of two firms is monopolist, each with a different bottleneck, and each firm requires the other’s bottleneck to produce its output Two local telephone companies, each customer subscribes only to one local telephone company, and each company requires the other’s network to complete calls Calls originate at A1, A2 and terminate at B1, B2. Termination charges at B1, B2 for calls from the rival network can be used to disadvantage and foreclose the rival network Example: New Zealand Problem in U.S. telecommunications solved by setting equal termination fees (reciprocity); unresolved in ATM, credit card, and other unregulated networks

170 Leveraging of market power across markets Various types of exclusionary arrangements Instruments: Technical standards Bundling and other pricing strategies Non-price discrimination strategies (raising rivals’ costs)

171 Limited effects of antitrust policy In markets with strong network effects, once few firms are in operation, the addition of new competitors, even under free entry, does not change the market structure in any significant way Although eliminating barriers to entry can encourage competition, the resulting competition may not significantly affect market structure In markets with strong network effects, antitrust authorities may not be able to significantly affect market structure by eliminating barriers to entry

172 Leveraging Example In the middle 1980s, Nintendo refused to allow third party games (software) to play on its game console (hardware) unless the software manufacturers agreed not to write a similar game for two years for competing game systems Nintendo used the dominance of the game market at the time to coerce developers to write software just for its platform, and thereby to increase the value of the Nintendo virtual network (of hardware and software) Practice stopped under threat from DOJ

173 Issues in “after-markets” where consumers are “locked-in” in a durable good or service arises out of commitments of durable nature Examples refusal of Kodak to supply to repair companies replacement parts for Kodak photocopiers lack of email address portability for ISPs lack of number portability for wireless phones long after it was feasible

174 Example from computing industry: subsidizing complementary goods Firm A chooses to make its product incompatible with others Firm A subsidizes firms that produce complementary goods Alternatively, firm A subsidizes its division that sells complementary goods As a result The value of firm A’s product increase The entry hurdle of firm A’s rivals increases Possible creation of market power, but action also has pro-competitive justification

175 Incompatibility is a necessary condition for possible creation of market power Key to increasing social welfare: public standards, compatibility But, it is very difficult for US antitrust authorities to intervene and/or define standards Different in the EU which is trying to impose compatibility between Microsoft and Sun servers (MS and Linux servers are compatible) Imposing compatibility may reduce incentives to innovate The “dynamically incorrect” standard may be chosen If forced to choose a single standard, the FCC would have chosen TDMA or GSM at the first PCS auctions CDMA proved more efficient later on Impose compatibility?

176 Dynamic efficiency issues Static efficiency may lack in dynamic efficiency Possibility of a lock-in to a technology which, when decisions taken in every period, looks optimal given past decisions, but is sub-optimal if decisions are delayed and all the decisions are taken at once Lock-in may occur as a consequence of the race to be a dominant firm in a network industry

177 Innovation issues Efficiency and intensity of innovation in monopoly compared to competition and oligopoly is an open question in economics

Regulation? 178

179 When should regulation be used? Regulation it is best suited for industries with well defined and not changing products and services Regulation is not well suited in industries with rapid technological change and frequently changing product definitions Regulation can be used by the regulated companies to keep prices relatively high, as exemplified by telecommunications regulation Often regulators are very close to the interests of the regulated parties rather than to the interests of the public Often regulators are not well informed about key variables as well as changes in the industry

180 When should regulation be used? Regulators at both the state and federal levels are under pressure and influence by both the executive and the legislative part of government, and cannot be as impartial as a court There is a tendency for regulators to expand their reach into related and new markets These drawbacks can create significant surplus loss due to regulation However, regulatory rules can and should be used effectively and appropriately in cases of dominance or monopolization of essential network bottlenecks to assure that firms do not leverage their monopoly to adjacent markets that prices are not too high

Week 8(a) Taxicabs; (b) Electric cars181

(a) Taxi cabs as a two-sided networkTraditionally taxicabs (yellow cabs) are hailed on the streettheir prices are regulatednumber of licenses is regulated; no entry allowedlack of entry sometimes leads to very high value of licenses (over $1mil in NYC)price of license is a fixed cost to the driver182

Parallel marketsPre-arranged ridesTown cars “Black cars”Liveries (in NYC above 96th Street and outside Manhattan)Each fleet uses incompatible dispatch systems183

New entrants, direct substitute for taxicabsLocations of cabs online using an app Customers “hail” cabs with the app based on location/distanceAdvantage: “see” free cabs further than the usual 2-3 blocksNo price regulationUber charges according to demand/supply imbalanceRegulatory issues at some municipalities184

(b) Electric cars as a two-sided networkNetwork issue because they need a network of fuel stations AlternativesFast charges at gas stations and homeReplaceable batteries at gas stationsPrecedent in electric taxicab fleets in New York and PhiladelphiaCars purchased from GE without a battery and Hartford Electric Light Company provided batteries and replacement 185

Week 8-9: (a) Payment Systems in US and EU; (b) mobile banking in Africa; Nicholas Economides (2009), Competition Policy Issues in the Consumer Payments Industry, in Robert E. Litan and Martin Neil Baily, eds., Moving Money: The Future of Consumer Payment Brookings Institution. Alan S. Frankel and Allan L. Shampine, The Economic Effects of Interchange Fees, Antitrust Law Journal 73, no. 3: 627–73. US DOJ press release, Settlement with Visa and MasterCard; United States v. American Express. Nicholas Economides and with David Henriques (2012), To Surcharge or Not to Surcharge? A Two-Sided Market Perspective of the No-Surcharge Rule, NET Institute Working paper #11-03, August 2011 186

187 (a) Payment Systems, Electronic Transaction Facilitation in US & EU Bank cards facilitate transactions between merchants and consumers Some cards also offer credit The market for facilitation of transactions is dominated by the card networks Visa and MasterCard Stand-alone cards American Express and Discover smaller Visa 42%, MasterCard 29%, American Express 24%, Discover 5% (US market shares) Card networks collect significant fees from merchants to facilitate transactions Cards charge (non-credit-related) fees (primarily to merchants) significantly above cost Some estimate total cost only 13-15% of revenue Fees $30-48 billion per year in U.S.

188 Three-parties Setup (American Express, Discover) American Express Merchant Consumer Goods worth $100 $97 $100

189 Four-parties Setup (MasterCard, Visa) Acquiring Bank Issuing Bank Merchant Consumer Goods worth $100 $98 $98.5 $100 Interchange fee Visa network

190 Three Markets in Sequence in Four-parties Setup Acquiring Bank Issuing Bank Merchant Consumer Goods worth $100 Market I Interchange fee Visa network Market II Market III Market III is generally considered effectively competitive, but since the interchange fee is the marginal cost of the acquiring bank, the interchange fee is a floor to the fees merchants pay In Market II, the network sets the maximum interchange fee Practically no bank in the network deviates from it No market determination of the interchange fee in bilateral markets between an acquiring and an issuing bank Issuing banks have market power in Market I but manage to make merchants pay through markets II and III rather than make card holders pay them directly

191 Credit Card Networks Have High Price-to-cost Markups Despite Non-Dominant-Firm Market Shares Very significant markup of price above cost Very unlikely that consumers receive from card networks anything approaching the fee levels charged to merchants Comparable with profit rates of Microsoft and Intel But these have an almost monopoly market share How does Visa with 42% market share and MasterCard with 29% market share achieve such high markups and market power?

192 Networks make sure that the consumers do not face directly the cost of their transactions so that they cannot choose to use the lowest fee card the merchants cannot charge different prices to reflect the card fees if consumers use different cards and different cards have different fees

193 The card networks Impose contractual obligations on merchants that do not allow merchants to respond to differences in fees across cards do not allow the card holders to choose which card to use based on the cost the transaction imposed on the merchant (since the merchant is not allowed to pass this along) Effects Card transactions are subsidized by cash transactions High cost card transactions are subsidized by low cost card transactions Significant market distortion

194 Instruments Used by Card Networks to Restrict Competition and Achieve High Fees (1) No surcharge rule (Visa contract, 2008) A merchant can charge the same for a Visa transaction as for cash If a merchant offers a discount for cash compared to Visa, he cannot offer the same discount to a “comparable card” (MasterCard) If a merchant offers a discount to a “comparable card,” he has to offer the same discount to Visa (“most favored customer” rule) Effect of the rule: the merchant cannot offer better terms to customers who buy with MasterCard than with Visa (although it may make sense to do so if MasterCard’s fees are lower) No price flexibility allowed in the merchant’s pricing As if Coca Cola imposed the requirement that a can of Pepsi be sold at the same price as a “comparable” can of Coke Only option for the merchant is not to accept a network card if he does not like the pricing

195 Instruments Used by Card Networks to Restrict Competition and Achieve High Fees (2) No discrimination rule MasterCard: “Merchants may not engage in acceptance practices or procedures that discriminate against, or discourage the use of, MasterCard cards in favor of any other card brand ....”

196 Instruments Used by Card Networks to Restrict Competition and Achieve High Fees (3) Card networks relied on a most-favored-customer (“MFC”) rule so that all prices among “comparable network cards” were the same IO theory has established that most-favored-customer rules can be used to facilitate price increases to collusive levels, Salop (1986) Reason: Under MFC rules, a firm will lose more revenue if it cuts price to a customer than in the absence of MFC rules

197 High fees are shielded through the “honor all cards” rule Debit networks (typically with PIN verification) offer lower merchant fees Debit cards of MasterCard and Visa also offer much lower fees than the signature-based credit cards To avoid loss of profits in credit cards, the networks impose the requirement that if a merchant accepts a Visa (say debit) card, he also has to accept all Visa cards (part of “honor all cards” rule)

198 Networks imposed the Honor all cards rule A merchant accepting a Visa debit card issued by Citibank also has to accept Visa Debit cards issued by any other Visa network bank Any Visa products, such as Visa credit cards (tying) Visa's rules stated that “[t]he Merchant shall promptly honor all valid Visa cards when properly presented as payment ....” Anti-competitive consequences of tying What if Microsoft said that if your corporation buys Windows, it also has to buy MS-Office Or if you buy Dell servers you also have to buy Dell laptops?

199 Effects of old equilibrium Card transactions are subsidized by cash transactions High cost card transactions are subsidized by low cost card transactions Networks have incentives to keep increasing interchange fees to attract more issuers

200 Proposed Changes in the Market Between Merchants and Acquiring Banks (Market III) Allow for competition between branded networks Abolish the “no surcharge” rule Abolish the “no discrimination” rule Abolish the “honor all cards” rule

201 Proposed Changes in the Market Between Acquiring and Issuing Banks (Market II) Allow the market between issuers and acquirers to determine the interchange fee Bilateral negotiations between pairs of banks (issuer and acquirer) Start from a par position (zero interchange fee)

202 Hold-out problem An issuer with very valuable potential transactions demands monopoly fee from acquirers Not a problem if monopoly power of issuer is legitimate Isolated problem if the proposal is adopted rather than market-wide problem under present regime Competition among issuers for these valuable customers solves the problem in the long run as these customers get more than one card

203 Effects of Proposed Changes Increased inter-network and intra-network competition likely to result in lower transaction facilitation fees

204 Australian Experience The Reserve Bank of Australia reduced interchange fees for credit cards in Australia from 0.95% to 0.55% in 2003 and to 0.50% in November 2006 Allowed surcharging Results: merchant fees fell even more than the interchange fees the overall cost to the economy of facilitating transactions fell

205 More radical changes Separating authentication from payment Large merchants eliminate acquirers?

In Sept. 2010, DOJ sued Visa, MasterCard and Amex over the no-surcharge rule Visa and MasterCard settled, accepting to abolish the no-surcharge and no discrimination clauses of their contractsAmex said it did nothing wrong; suit U.S. v. Amex continuesImpact on merchants who do not take Amex? 206

EU actions, July 2013EU payment market is worth 130bn eurosEU wants to cap interchange fees to a maximum of 0.3% of a transactionDebit card fees to 0.2% of a transactionSee http://www.bbc.co.uk/news/business-23431543 207

(b) Mobile banking in Africa and other developing areas Read: Seeking Fertile Grounds for Mobile Money208

209 Many households do not have traditional bank accounts A new bank account can be created and attached to a mobile phone number and therefore to a mobile telecom network Each mobile network creates its own money Mobile banking networks are incompatible When are two mobile telecom companies, A, B, in a country , there are three types of money: State (S-) money, A- m-money , and B- m-money

210 Conversion from S- money to A- or B- m-money (“cashing in) is free Conversion from A- or B- m-money to S-money (“cashing out”) is expensive: fee 2.5% for an average transaction Transfer across networks, from A- money to B- money is even more expensive, fee 2.8% for average transactions Transfer within a network has a low fee, 0.3% for average transactions Substantial surplus losses because of de facto lack of interoperability De facto lack of interoperability makes exchange rates crucial. In Tanzania:

In Tanzania 35% of households have at least one m-money account three major m-banking networksVodacom (Vodafone), 53% market share in m-moneyTigo, 18%Airtel, 13%Calling and transferring money across networks is expensive, many consumers have a different phone or a different SIM for each networkTigo advertises phones that take multiple SIMs 211

M-money is most frequently used to send or receive remittancesNon-remittance transfers are infrequent 14% of all households made or received a non-remittance payment in the past six months using any type of cash delivery, including m-moneyThe most common types of payments included school fees, government fees and taxes, utility bills, and salaries 212

Despite the zero interest rateHalf of consumers with m-money accounts use their account as a savings account 21% Vodacom M-Pesa users and 12% of users of Tigo & Airtel use m-money for business transactions 213

Cashing in and cashing out of m-moneyis done through a network of fixed and roaming agents that act as ATM machines This is the main cost of the banking networks 214

Data3 months of banking transactions from Tigo Tanzania, includingcash-in, cash-out, recharge mobile, transfer, check balanceBalance, operation size, GPS location, feeDuring the period, there was an unanticipated price change in transfer and cash-out fees 215

Tanzania 216

Frequency of transfers 217

Transfer network 218

Cash is a close substitute for short distance transfers. Elasticity of demand for m-transfers decreases with distance. 219

Probability of an m-transfer increases with distance 220

Summary of resultsLong distance transfers more inelastic, 1.25 vs. 1.55 Transaction taxes 0.25% transaction tax lowers the propensity to transfer by 46%10% tax on the transfer fee lowers the propensity of transfers by 13%Taxes are extracted from long distance remittancesregressive tax Interoperability with zero cash-out fee would increase the propensity for m- transfers by 1%221

Week 10: The Digital Books MarketMarket Structure Distribution model of digital goodsDOJ suit and EU statement of objections against Apple and five publishersUS ComplaintSettlement of publishersCourt decision on Apple 222

History and Market Structure1992 Sony sells Data Discman, an eBook reader for CDs 2000 Microsoft launches Microsoft Reader, works in Windows2007 Amazon launches Kindle2010 Apple launches iPad2011 Barnes and Noble launches Nook223

Book sales revenue 2012, eBooks surpass hardcover 224

Retail price for ebooks used to be set by retailer (Amazon, etc.) Publisher collects price P per eBookRetailer sets price R to usersPrice R for most ebooks > P, but sometimes R < PBefore the iPad came to market, Amazon sold “NYT bestsellers” for R = $9.99, below P225

Publishers did not like low R Even though they still collected P > RWhy?Because most of the book sales were of hard copies, and the hard copy price Ph was much larger than RPublishers were worried that they may have to reduce price Ph (say $35) in the presence of an eBook price of $9.99Before the iPad came to market, in Dec. 2009 publishers retaliated against Amazon by withholding availability of “new releases” as ebooks for a few months after hard copies of new books were being sold, a practice called “windowing”But windowing encouraged piracy 226

In the months before the launch of the iPadApple was in intense negotiations with publishers attempting to convince them To adopt the “agency pricing model”To impose this model to all retailers, including AmazonTo allow for ebooks of NYT bestsellers to become available at the same time as hard copiesTo cut the price of eBook NYT bestsellers 227

Once the iPad entered the market the 5 major publishers and Apple agreed to move to the “agency pricing model” on the condition that it would apply to Amazon tooAgency pricing model:Retail price R is set by publisherRetailer keeps 30% of the retail price, i.e., P = (0.7)RSimilar to the way Apple sold third party software for the iPad through its virtual store 228

Suit of DOJ against Apple and publishers for conspiring to set eBook prices (April 2012). Defendants: Hachette Book Group, Inc.HarperCollins Publishers L.L.C.Simon & Schuster, Inc.MacmillanPenguin groupThe above settled; Remaining defendantApple 229

Apple CEO Steve Jobs to Publisher Defendants:“We’ll go to [an] agency model, where you set the price, and we get our 30%, and yes, the customer pays a little more, but that’s what you want anyway.” (p. 4 of US complaint) 230

Effects of the conspiracyThe increases at Amazon within roughly two weeks of moving to agency amounted to an average per unit eBook retail price increase of 14.2% for their New Releases, 42.7% for their NYT Bestsellers, and 18.6% across all of the Publisher Defendants’ e-booksCourt decision on Apple, page 94231

Prices increased as a result of the conspiracy 232

Ebook prices before and after the conspiracy agreement 233

Strength of DOJ’s caseConspiracy (Section 1 of Sherman Act), no need to show anti-competitive effects) Weakness of the caseApple can claim it was imposed to it by the publishersDOJ could have brought criminal charges but did notShows that DOJ did not believe it could win a criminal case 234

Settlement approved 9/7/12Ends conspiracy among publishers Allows eBook retailers to set price (as they used to)Settling publishers agreedto refrain for two years from entering into contracts containing retail price restrictions and price commitment mechanismsto stop communicating competitively sensitive information to competitorsnot retaliate against retailers that exercise discounting authoritynot to fix terms or prices with competitors for the provision of e-booksAGs of various states get $69 million damages from publishers 235

Similar case in EU settledOn 9/7/12, Apple and four major publishers (all US defendants except Penguin) have offered to allow retailers such as Amazon to sell e-books at a discount for two years On 11/6/12, Reuters reports that EU is about to settle on these terms 236

In July 2013, US District Court decision finds Apple liableof conspiracy to restrain trade in violation of Section 1 of the Sherman Act Civil suits to followNeed additional study of prices after the end of the conspiracy237

DOJ proposal on remedies: Apple is required to end its existing eBook agreements with five major publishers and sign no new price-setting distribution contracts for five years allowed to continue to sell ebooksprohibited from retaliating against publishers for refusing to sell ebooks under terms approved by Applerequired to allow other eBook sellers, such as Amazon and Barnes & Noble, to give more prominent play to their eBook stores on Apple devices, by allowing them to provide links from their eBook apps to their eBook storesprohibited from entering into agreements with suppliers of ebooks, music, movies, television shows or other content that are likely to increase the prices at which Apple's competitor retailers may sell that content See http :// www.computerworld.com/s/article/9241316/DOJ_AGs_propose_remedies_for_Apple_in_e_book_price_fixing?source=CTWNLE_nlt_pm_2013-08-02 238

Week 11: Two-sided Pricing & Network NeutralityGeneral theory Application to network neutralityPublic policyReferences:Economides & Tag, “Network Neutrality on the Internet: A Two-sided Market Analysis,” Information Economics and Policy (2012), at http://www.stern.nyu.edu/networks/Economides_Tag_Net_Neutrality.pdfEconomides and Hermalin (2012), “The Economics of Network Neutrality,” forthcoming, Rand Journal of Economics, at http://www.stern.nyu.edu/networks/Economides-Hermalin_Economics_of_Network_Neutrality.pdf 239

240 In networks, price discrimination schemes can be complex (1) Because a network may have different degrees of market power on different sides of the market A firm that controls a proprietary platform sets a price strategically for its end-user products and collects a fee for complementary products to its platform sold by other firms Sony controlling its game console sets a price for the game console and charges royalties to developers of games Microsoft sells Windows to end-users provides application developers with information and resources; embeds subroutines in Windows useful to application developers makes money from licensing application development tools and support

241 Prices and fees in two-sided markets p 1 s 1 p 0 s 2 s 3 p 2 p 3 p 0 user platform price p 1 , p 2 , p 3 user application prices s 1 , s 2 , s 3 fees to platform (or subsidies to applications)

242 In networks, price discrimination schemes can be complex (2) Since the creation of the Internet, there was no price (or other) discrimination based on what service or what application the bits came from (so called “net neutrality”) Now AT&T, Verizon and cable TV networks advocate price discrimination based on which application and on which provider the bits come from AT&T, Verizon and cable TV networks propose to kill net neutrality by charging both the DSL subscriber and the application provider (such as Google) where the bits originate, even when the application provider is not directly connected to AT&T or Verizon (i.e., Google’s ISP is not AT&T or Verizon) Note that the proposal is to impose price discrimination on the provider side of the market and not on the subscriber (i.e., it is a version of two-sided pricing)

243 Two-sided Pricing & Network Neutrality Internet Backbone AT&T (Access Network) Netflix Google Residential Customers R : subscription price s : AT&T’s fee to content providers or many fees s1, s2, …, sn ISP ISP p

244 Interview with Ed Whitacre BusinessWeek November 7, 2005 How concerned are you about Internet upstarts like Google, MSN, Vonage, and others? “How do you think they're going to get to customers? Through a broadband pipe. Cable companies have them. We have them. Now what they would like to do is use my pipes free, but I ain't going to let them do that because we have spent this capital and we have to have a return on it. So there's going to have to be some mechanism for these people who use these pipes to pay for the portion they're using. Why should they be allowed to use my pipes? The Internet can't be free in that sense, because we and the cable companies have made an investment and for a Google or Yahoo! or Vonage or anybody to expect to use these pipes [for] free is nuts!”

245 But Both Sides Pay for “Transit” on the Internet All hosts on the Internet pay according to bandwidth use: there is no “free lunch” on the Internet AT&T, Verizon, and others are paid by ISPs according to bandwidth use Actually Internet backbones are paid twice for any transmission, by the originator of traffic and by the terminator of traffic (through their respective ISPs)

246 Varying Levels Of “Network Neutrality,” from Strictest to Weakest Referring to pricing to the “other side” of the consumer market (that is to content and applications providers): 1. Absolute non-discrimination: no quality of service variations offered for money or for free 2. Varying quality of service offered according to type of info. packet but no fees are charged 3. Tiered service allowed but each tier is offered at the same price to all; no exclusivity or identity-based discrimination 4. Identity-based discrimination allowed 5. Exclusivity allowed

247 Obama’s statement on NN, at http://www.barackobama.com/pdf/issues/technology/Fact_Sheet_Innovation_and_Technology.pdf Barack Obama strongly supports the principle of network neutrality to preserve the benefits of open competition on the Internet. … Because most Americans only have a choice of only one or two broadband carriers, carriers are tempted to impose a toll charge on content and services, discriminating against websites that are unwilling to pay for equal treatment. This could create a two tier Internet in which websites with the best relationships with network providers can get the fastest access to consumers, while all competing websites remain in a slower lane. Such a result would threaten innovation, the open tradition and architecture of the Internet, and competition among content and backbone providers. It would also threaten the equality of speech through which the Internet has begun to transform American political and cultural discourse. Barack Obama supports the basic principle that network providers should not be allowed to charge fees to privilege the content or applications of some web sites and Internet applications over others. This principle will ensure that the new competitors, especially small or non-profit speakers, have the same opportunity as incumbents to innovate on the Internet and to reach large audiences.

FCC’s NPRM (2009) (1): “Subject to reasonable network management, a provider of broadband Internet access service may not 1. prevent any of its users from sending or receiving the lawful content of the user’s choice over the Internet.2. prevent any of its users from running the lawful applications or using the lawful services of the user’s choice.3. prevent any of its users from connecting to and using on its network the user’s choice of lawful devices that do not harm the network.4. deprive any of its users of the user’s entitlement to competition among network providers, application providers, service providers, and content providers.” 248

FCC’s NPRM (2009) (2): “Subject to reasonable network management, a provider of broadband Internet access service must treat lawful content, applications, and services in a nondiscriminatory mannerWe understand the term “nondiscriminatory” to mean that a broadband Internet access service provider may not charge a content, application, or service provider for enhanced or prioritized access to the subscribers of the broadband Internet access service provider … We propose that this rule would not prevent a broadband Internet access service provider from charging subscribers different prices for different services.” 249

FCC Rule (December 2010) 1. Transparency: Fixed and mobile broadband providers must disclose the network management practices, performance characteristics, and terms and conditions of their broadband services.2. No blocking: Fixed broadband providers may not block lawful content, applications, services, or non-harmful devices; mobile broadband providers may not block lawful websites, or block applications that compete with their voice or video telephony services.3. No unreasonable discrimination: Fixed broadband providers may not unreasonably discriminate in transmitting lawful network traffic.Even though this regulation is weak, Verizon appealed to stop it. In Jan. 2014, Verizon won, and the FCC rules were nullified. 250

EU: Directive 2009/140/Ec 25 November 2009 Commission Declaration On Net Neutrality The Commission attaches high importance to preserving the open and neutral character of the Internet, taking full account of the will of the co-legislators now to enshrine net neutrality as a policy objective and regulatory principle to be promoted by national regulatory authorities alongside the strengthening of related transparency requirements and the creation of safeguard powers for national regulatory authorities to prevent the degradation of services and the hindering or slowing down of traffic over public networks 251

EU: Directive 2009/140/Ec 25 November 2009, Consultation, and Kroes SpeechNational Regulatory Authorities are required to promote “the ability of end-users to access and distribute information or run applications and services of their choice”National Regulatory Authorities, after consulting the Commission, can set minimum quality of service requirementsStrong transparency to ensure consumers understand and get what they pay for 252

253 Residential ISPs have market power; also claim congestion Assuming no congestion, Economides and Tag (2009) and others showed that introducing a positive price to content providers is typically welfare-inferior to NN Economides & Tag, “Network Neutrality on the Internet: A Two-sided Market Analysis,” Information Economics and Policy (2012), at http://www.stern.nyu.edu/networks/Economides_Tag_Net_Neutrality.pdf ISPs claim that NN is not optimal with congestion Issue addressed in Economides and Hermalin (2012), “The Economics of Network Neutrality,” Rand Journal of Economics , at http://www.stern.nyu.edu/networks/Economides-Hermalin_Economics_of_Network_Neutrality.pdf

254 Six Consequences of Departure from Net Neutrality 1. Introduction on the Internet of two-sided pricing where a transmission company controlling some part of the Internet (here last mile access) will charge a fee to content or application firms “on other side” of the network Economides and Tag (2012) main result: Starting to charge a positive price “s” on the “other side” of the market is desirable to an access monopolist (or duopolists) but not desirable for society – more later in the presentation

255 Six Consequences of Departure from Net Neutrality 2. Introduction of prioritization which may enhance the arrival time of information packets that originate from paying content and application firms “on the other side,” and may also degrade the arrival time of information packets that originate from non-paying firms The present plans of access providers are to create a “special lane” for the information packets of the paying firms while restricting the lane of the non-payers without expanding total capacity By manipulating the size of the paying firms’ lane, the access provider can guarantee a difference in the arrival rates of packets originating from paying and non-paying firms, even if the actual improvement in arrival time for paying firms’ packets is not improved over net neutrality

256 Six Consequences of Departure from Net Neutrality 3. If the access providers choose to engage in “identity-based” discrimination, they can determine which one of the firms in an industry sector on the other side of the network, say in search, will get priority and therefore win This can easily be done by announcing that prioritization will be offered to only one of the search firms, for example the one that bids the highest Thus, the determination of the winner in search and other markets on the other side will be in hands of the access providers and not determined by innovative products or services on the other side This can create very significant distortions since the surplus “on the other side” of the Internet is a large multiple of the combined telecom and cable TV revenue from residential Internet access

257 Six Consequences of Departure from Net Neutrality 4. New firms with small capitalization (or those innovative firms that have not yet achieved significant penetration and revenues) will very likely not be the winners of the prioritization auction This is likely to reduce innovation. Network externalities arise because a typical subscriber can reach more subscribers in a larger network Under no net neutrality, access providers can limit the size and profitability of new firms on the “other side”

258 Six Consequences of Departure from Net Neutrality 5. The access networks can favor their own content and applications rather that those of independent firms Examples: independent VOIP, video This is likely to distort competition and reduce total surplus

259 Six Consequences of Departure from Net Neutrality 6. Since the Internet consists of a series of interconnected networks, any one of these, and not just the final consumer access ones, can, in principle, ask content and application providers for a fee This can result in multiple fees charged on a single transmission and lead to a significant reduction of trade on the Internet

260 Economides-Tag (2012) deals with two-sided pricing when there is no congestion In terms of potential welfare reduction because of the six effects discussed above, we model the case that has the least reduction in total surplus compared with net neutrality Even though we make the best possible case for total surplus to increase when departing from net neutrality (by not focusing on factors two to six that are likely to reduce surplus), we find that typically total surplus decreases, both in monopoly and duopoly when we depart from net neutrality

261 Stylized model: Money flows Platform (ISP) Consumers Content / Applications Providers s p R

Crucial QuestionsDoes the platform want to set s >0? Does the platform want to set s1, s2, etc.?Are consumers better/worse off at s > 0 vs. s = 0 Are consumers better/worse off at s1 > s2 > 0 vs. s = 0Is total surplus (TS = CS + Profits of ISP + Profits of Apps) higher with s > 0 than s = 0? With s1 > s2 > 0 vs. s = 0? 262

Model 1 (Economides and Tag, 2012) Network effects from the consumers to the apps and vice versaMonopolistic competition among appsMonopoly or duopoly ISPNo congestionHeterogeneous consumersHeterogeneous appsNetwork effect from consumers to apps “a”Network effect from apps to consumers “b” 263

264 Platform incentives in setting fee s M to other side of the market Platform Consumers a = value (network effect) of extra consumer to a content provider s M R M > 0 b = value (network effect) of extra content provider to a consumer a > b  s M > 0 Internet consumers, access platform, and content providers Game platform consumers, game platform, and games a < b  s M < 0 PC users, operating system, and applications Credit card issuing banks, credit card platform, and consumers Content Providers

265 Optimal One-sided Regulation in the Presence of Monopoly on the Other Side of the Market A regulator/planner setting a fee s to content providers expecting the platform monopolist to set his profit-maximizing subscription price p(s) maximizes the constrained total surplus function TS(p(s), s) and chooses a below-cost fee to content providers s***< 0 provided that both consumers and content providers are sufficiently differentiated Even paying the below-cost fee, the platform makes positive profits Similar results for duopoly ISPs

Economides-Hermalin assumes congestionExamines under what conditions NN is optimal What pricing does the ISP implement? 266

Model 2 (Economides and Hermalin, 2012) Assume congestion: If traffic exceeds the pipe’s capacity it gets delayed proportionatelyMonopoly ISPHomogeneous consumersHeterogeneous appsBelow: “welfare” = “total surplus TS”TS = CS + profits of ISP + profits of appsBecause the consumers are homogeneous, the ISP appropriates all CS 267

Define net neutrality as no division of the bandwidth Otherwise, bandwidth is divided in lanes, and each lane is allocated to one or more appsProvided that the monopolist ISP allows (sells to) all apps:Neutrality is weakly welfare-superior to any division of the bandwidth in which all segments are allocated a positive portion of the bandwidthNeutrality is strongly welfare-superior to any division that excludes some segments 268

269 A bandwidth allocation is welfare-superior if it carries more content Consider divisions of bandwidth in lanes Proposition 1 . Given two alternative divisions of the total bandwidth, one is welfare superior to the other if and only if it results in more content being carried in equilibrium than the other. Corollary 1 . Network neutrality is welfare superior to any division of the bandwidth if no division of the bandwidth leads to more content being sent in equilibrium.

270 Amount of content carried is a “sufficient statistic” for welfare and how bandwidth is allocated to content does not matter Content is analogous to income, and it has the maximum positive effect on welfare when it is allocated as the consumer sees fit without restrictions Restrictions (special lanes etc.) in its allocation restrict welfare [Reminiscent of Varian (1985,9) showing that third degree price discrimination increase welfare vis-a-vis uniform pricing only if it increases the total amount of goods sold]

271 Exclusion reduces welfare Proposition 2 . Suppose the ISP provides a common class of service, but excludes a positive measure of content providers. In the resulting equilibrium, TS is less than it would be were no content providers excluded.

Each user likes more applicationsWhen the consumer surplus is higher, the ISP (who appropriates consumer surplus) has an incentive to allow more appsEach app wants fewer (other) apps because more apps congest the network and delay deliveryThis externality among apps is not internalized in this model 272 Allowing the ISP to charge a positive fee to apps will exclude some apps

Welfare = Total Surplus = Consumer Surplus + Profits of ISP + Profits of Apps A necessary condition for welfare under price discrimination to exceed welfare is that the set of application providers served in equilibrium expandIf the ISP serves all application providers in equilibrium under linear pricing, then any non-trivial price discrimination scheme reduces welfare 273

The effect of prioritization on total surplusDepends on how the elasticity of demand with respect to transmission time varies with content type NN can maximize welfare if the elasticity of demand with respect to transmission time is invariant with respect to contentBut depending on how that elasticity varies with content type, to max. welfare you may want to prioritize or slow down content of high immediacy desire 274

Crucial issueis how many apps at the bottom are excluded when fee to apps is introduced i.e., depends on the distribution of app types Welfare can be higher or lower in price discrimination vs. linear pricing, depending on parametersOf course, higher welfare requires more total content sales 275

Dynamic Issues: investment in bandwidth ISP invests more in bandwidth with single positive fee to apps than with no feeWith single positive fee, ISP invests less in bandwidth compared to the welfare-maximizing amountISP invests more in bandwidth with price discrimination than with single positive fee to apps 276

Distinction between static and dynamic efficiency Example where θ (app types by desirability of immediacy) is distributed uniformly on [.05,1].Given a fixed bandwidth, compelling ISP to serve all application providers would raise welfare by 10%Cost of bandwidth κB (unrealistic technology)Unrestricted ISP will build 1.649 times as much bandwidth as it would were it constrainedGreater bandwidth does not fully overcome the static inefficiency, but, once the dynamic consequences are considered, welfare with zero fee is only 2% greater than were the ISP unrestricted 277

In some models (Choi and Kim, Rand Journal 2010), ISP can have higher profits if it restricts bandwidth Lower bandwidth can increase the difference between the price of a good in the fast lane vs. slow laneNot in Economides-Hermalin 278

279 Conclusions on net neutrality (1) Two-sided pricing is complex Identified conditions when prices are positive on both sides and when one side is subsidized Generally, social welfare diverges from platform profits On the Internet, net neutrality typically is welfare-maximizing

Conclusions on net neutrality (2)Under reasonable assumptions, welfare increases in the total amount of content carried Even in the presence of congestion, NN can maximize welfare if the elasticity of demand with respect to transmission time is invariant with respect to contentBut depending on that elasticity, to max. welfare you may want to prioritize or slow down content of high immediacy desireKilling NN can increase bandwidth investment and reduce the static allocative distortion 280

Week 12: Bottom line 281

282 Companies’ strategies that were successful in network industries 1. Vertical extension of the company 2. Discount pricing based on volume to take advantage of network effects 3. Discount pricing based on volume to take advantage of dominant position and disadvantage competitors 4. Subsidizing complementary goods 5. Control of bottlenecks 6. Exclusive contracts

283 1. Vertical extension of the company Example: Microsoft 92% market share in operating systems for PCs Over time, it added functions to the OS that used to be independent applications or middleware: Browser Windows Media Player Hard disk defragmenter Anti-spyware Anti-virus But MS does not go in hardware in PCs; is aware of its core competency

284 Advantages of extending the firm vertically Offensive advantage Takes away value from complementary goods and adds value to own product Defensive advantage Avoids complementary goods firms creating a challenge to own product Netscape browser + Java might challenge Windows

285 Disadvantages of extending the firm vertically Strategy may be illegal as in the Microsoft cases in the US and EU But strategy very likely legal if market share is < 50% in US and < 40% in EU Vertical extension could distract from the core competency of the firm

286 Success? Very successful for Microsoft against Netscape which saw its share going from 100% to 0% But, so-and-so in audio players

287 2. Discount pricing based on volume to take advantage of network effects Example: Cantor Fitzgerald in the secondary market for US government bonds; had 70+% share Offered per unit pricing very significantly above marginal cost to all traders except Salomon Brothers Salomon was offered marginal cost pricing plus a fixed fee Why? To get the very high liquidity of Salomon that had 40+% in primary market, and internalize the network effects

288 Success? Very successful for Cantor Kept dominant position despite an inefficient trading platform (broker matching vs. electronic trading) Primary dealers had to set up their own exchanges as a threat to constrain the per-unit pricing of Cantor

289 3. Discount pricing based on volume to take advantage of dominant position and disadvantage competitors Example: Microsoft’s “per processor” pricing before 1995 Offered Windows at a per unit price, say $30 Also offered Windows to a PC manufacturer, say Compaq, that produced say 1 mil. units, at a flat price of $24 mil. with the right to install in all units For Compaq, the last 200,000 units effectively had zero price Compaq had a strong incentive not to buy from competitors (IBM)

290 Advantages When marginal cost is very low, this strategy can rapidly increase market share (as it did for Microsoft) Great success for Microsoft that marginalized DOS competitor DR-DOS; increased the market share of Windows (that faced IBM’s OS2) to 90+%

291 Disadvantages The strategy is likely to be illegal for a dominant firm in the US, EU, Japan, and Korea In 1995 (before the big antitrust suit) Microsoft agreed with USDOJ to stop using this strategy But, it is a legal strategy for non-dominant firms

292 4. Subsidizing complementary goods Example: Microsoft It chooses to make its product (OS) incompatible with others It subsidizes firms that produce complementary goods (by including in Windows features that are useful to application developers but not to users) Alternatively, MS subsidizes its division that sells complementary goods (Office) As a result The value of MS’s product increases The entry hurdle of MS’s rivals increases

293 Advantages of subsidizing complementary goods Offensive advantage The value of own product increases Defensive advantage The entry hurdle of rivals increases It can make a platform dominant Huge mistake of Apple not to adopt this strategy when Bill Gates pushed Apple to adopt it (before the creation of Windows) Refusal by Apple prompted MS to create Windows and made MS the key player in PCs

294 Disadvantages of subsidizing complementary goods It costs money that may not be recoverable for some time However, as Microsoft showed, subsidizing complementary goods can create dominance in the long run with very large benefits

295 5. Control of bottlenecks Is crucial for creation and maintenance of dominant market position Sometimes innovation can eliminate the bottleneck Example: software services can eliminate the bottleneck of OSs

296 Old software model with OS bottleneck Windows OS Windows Application Software Windows dominant Application needs OS-specific APIs Apple OS X Mac Application Software Linux OS Linux Application Software

297 New software service model eliminates/reduces the bottleneck Browser Windows OS Application as a Service Windows dominant Open standards OS market structure irrelevant for service application market Apple OS Linux OS Based on browser’s open standards

298 6. Exclusive contracts Example: Microsoft Exclusive contracts with AOL and other ISPs on adoption of Internet Explorer Successful but illegal for a dominant firm

299 Some Big Upheavals in Network Industries 1. Google in Internet search/advertising 2. iPod big success in digital audio 3. VHS killing Betamax 4. Internet Explorer displacing Netscape 5. Alternating current displacing direct current in electricity

300 Being first is no guarantee of long run success Examples Google in Internet search Arrived very late Used a different algorithm for search Based its revenue on advertising Apple’s iPod Also arrived very late Hardware-software combination More liberal contract on legal copies (at the time)

301 Betamax/VHS: Sony’s strategic error VHS video recorder (JVC/Matsushita) Came later than Betamax (Sony) Was widely licensed; low price Betamax was not widely licensed; high price Much bigger network effects of VHS After 5 years, Betamax withdraws from the USA Strategic error of Sony Originally video recorders were used for time-shifting (like Tivo) – there were no network effects Only later movies for rental appeared -- crucial complementary good creating network effects Sony’s managers missed the transition …

302 Internet Explorer displaces Netscape Originally Netscape (based on Mosaic of the Univ. of Illinois) was dominant with 90+% market share Microsoft makes a huge effort to write better browser from scratch IE3 was significantly better than Netscape Microsoft uses exclusive contracts and bundling in Windows to boost IE’s market share

303 Alternating current displacing direct current in electricity Originally electricity generation and distribution was developed as direct current (DC) by Edison Significant municipal networks (New York City, Philadelphia) were created based on DC Light bulbs last much longer on DC Westinghouse pushed AC because its motors run much more efficiently on AC AC won because of the efficiency of its long distance transportation (Niagara Falls to NYC)

304 Bottom line (1) 1. Incompatibility is key 2. Under incompatibility, “winner-takes-most” 3. Crucial to have the top market share 4. Competition for the market more important than competition in the market ; make early sacrifices 5. Extend the firm vertically without going outside core competences

305 Bottom line (2) 6. Use price discrimination to a. take advantage of network effects b. take advantage of or create dominant position c. disadvantage competitors 7. Subsidize complementary goods to create a dominant platform 8. Try to keep control of bottlenecks; create new bottlenecks if possible 9. May use exclusive contracts if legal

Weeks 12: Antitrust in Networks Industries3 Microsoft cases The Microsoft Antitrust Case For MBA Students NYU conference on US v. MS, including streaming video of all presentations featuring (among others) Assistant Attorney General for Antitrust Doug Melamed, NY Assistant Attorney General for Antitrust Harry First, Microsoft counsel Rick Rule and Former Solicitor General Boyden Gray Discussion on US v. MS on PBS TV with host Jim Goodale, Prof. Nicholas Economides, and Prof. Eleanor Fox, in streaming video, first broadcast on November 16, 2000 Story in Wired magazine on Microsoft’s proposal for Apple allow clones Nicholas Economides & Ioannis Lianos, A Critical Appraisal of Remedies in the EU Microsoft Cases, Columbia Business Law Review 2010/2:346-420 (2010).Nicholas Economides & Ioannis Lianos, The Quest for Appropriate Remedies in the Microsoft Antitrust EU Cases: A Comparative Appraisal, Microsoft on Trial: Legal and Economic Analysis of a Transatlantic Antitrust Case , Luca Rubini (ed.). Edward Elgar (2010). Nicholas Economides & Ioannis Lianos, The Elusive Antitrust Standard on Bundling in Europe and in the United States at the Aftermath of the Microsoft Cases , Antitrust Law Journal 76/3:483-567 (2009 ). Investigations of Google 306

Microsoft AntitrustCalendar: Early Fights 1991-93: FTC investigates MS twice, but does takes action1994-95: DOJ’s investigation ends in a 1995 settlementKey provisions:1. Microsoft agrees to end “per-processor” (zero marginal price) contracts with OEMs but can use unrestricted quantity discounts2. “Microsoft shall not enter into any License Agreement in which the terms of that agreement are expressly or impliedly conditioned upon the licensing of any other Covered Product, Operating System Software product or other product (provided, however, that this provision in and of itself shall not be construed to prohibit Microsoft from developing integrated products); or the OEM not licensing, purchasing, using or distributing any non-Microsoft product.”That is, no product bundling allowed by contract, but Microsoft can keep expanding the functions of its products, including Windows 307

308 1997: Senator Orin Hatch (R-Utah) holds congressional hearings on Microsoft featuring Gates, Barksdale, Dell. Sen. Hatch takes the position that if present antitrust law cannot deal with Microsoft, Congress should change or enhance the antitrust laws Sun, Oracle, IBM, Netscape, and Novell form a loose coalition lobbying for antitrust action 1997: DOJ alleges anti-competitive bundling of IE with Windows (violation of 1995 decree) Dec. 1997: Judge Thomas Penfield Jackson issues preliminary injunction barring bundling of IE with Windows 98 May 12, 1998: Court of Appeals (DC Circuit) decides that 1995 decree doesn’t apply to Windows 98, which was shipped with integrated IE May 18, 1998: DOJ and 20 states and the District of Columbia file the main antitrust case Oct. 1998 – June 1999: Microsoft trial takes place with an accelerated schedule Main U.S. Fight (1)

309 Main (U.S.) Fight (2) Nov. 5, 1999: Judge Jackson issues “findings of fact,” siding very strongly with the plaintiffs Dec. 1999: Prominent antitrust scholar, Judge Richard Posner appointed mediator for settlement discussions April 1, 2000: Settlement talks break down after States hold out in proposed agreement. April 3, 2000: Judge Jackson issues “conclusions of law” June 7, 2000: Judge Jackson orders breakup of Microsoft in two companies February 27, 2001: DC appeals court hears appeal June 28, 2001: DC appeals court reverses breakup September 6, 2001: DOJ seeks quick settlement without breakup November 2, 2001: DOJ and Microsoft propose settlement; nine states settle and nine do not March 18, 2002: Nine litigating states start remedies trial in front of Judge Colleen Kollar-Kotelly November 1, 2002: Judge Colleen Kollar-Kotelly rules that the proposed settlement serves the public's interest and rejects positions of litigating states November 29, 2002: All states except Massachusetts and West Virginia accept the settlement; eventually, both Massachusetts and West Virginia agree with the settlement

310 Microsoft’s business then Software Operating systems for PC (Windows 95, 98, NT, 2000) Operating systems for local network and Internet servers (Windows NT, 2000) “Back-office” products for network and Internet servers Internet Clients Internet Servers Desktop applications (Office, Word, Excel, Access, PowerPoint, MS-Money, etc.) Games Programming languages (Visual Basic, Java) Hardware Mice, keyboards Services Internet service (MSN, WebTV) Internet content (MSN) Product support

The allegations.Microsoft accused of: 1. Monopolization of the market for operating systems (“OSs”) for PCs; (¶ 2, Sherman Act)2. Anti-competitive contractual arrangements with various vendors of related goods such as with computer manufacturers and Internet Service Providers (“ISPs”) and other actions taken to preserve and enhance its monopoly; (¶ 2, Sherman Act)3. Attempting monopolize the market for Internet browsers (but failing to succeed); (¶ 2, Sherman Act)4. Anti-competitive bundling of the Internet Explorer (“IE”), with the Windows operating systems; (¶ 1, Sherman Act) 311

312 “Monopolization” under ¶ 2 of the Sherman Act is illegal if the offender took anti-competitive actions to acquire, preserve, or enhance its monopoly Sherman Act ¶ 2: “Every person who shall monopolize, or attempt to monopolize, or combine or conspire with any other person or persons, to monopolize any part of the trade or commerce among the several States, or with foreign nations, shall be deemed guilty of a felony.” Modern interpretation: For “monopolization,” plaintiffs have to prove that the defendant 1. Possesses market power 2. Willfully acquired or maintained this monopoly power as distinguished from acquisition through a superior product, business acumen, or historical accident

“Attempting to monopolize” is illegal (Sherman Act, ¶ 2) Bundling, and, more generally, price discrimination could be illegal if it has anti-competitive consequencesExclusionary contracts could be illegal if they have anti-competitive effectsTo prove “attempting to monopolize” (under Sherman Act ¶ 2), plaintiffs have to prove that the defendant 1. Engaged in predatory or anti-competitive conduct 2. with specific intent to monopolize 3. and that there was a “dangerous probability” that the defendant would succeed in achieving monopoly power 313

Findings of fact and conclusions of law The judge’s “findings of fact” (Nov. 1999) and “conclusions of law” (April 2000) found for the plaintiffs (US DOJ and 19 States) in almost all the allegations against MSThe judge found:Microsoft has a monopoly in the PC operating systems market (for Intel-based computers) “where it enjoys a large and stable market share”Microsoft used its monopoly power in the PC operating systems market and harmed competitorsMicrosoft hobbled the innovation processMicrosoft actions harmed consumersVarious Microsoft contracts had anti-competitive implications, but MS is not guilty of anti-competitive exclusive dealing contracts hindering the distribution of Netscape 314

315 On June 7, 2000, after an extremely short hearing, Judge Jackson issues his remedies decision: split Microsoft into two companies, one for operating systems (Windows 98, NT, and 2000), and one for all the rest, including applications (MS-Office, etc.) impose business conduct restrictions

316 Problem: Low price of OS; if Microsoft is able to exercise monopoly power, why does it not exercise it through price? If the price of PC hardware is pH and the demand elasticity is |ε|, the monopoly price of Windows is pW = pH/(| ε| - 1) If pH = $1,800 and |ε| = 2, pW = $1,800, while the actual price to OEMs was $40-60 Requires a very large demand elasticity of |ε| = 31 to get a monopoly price of pW = $60 Judge decided that MS has monopoly power in the OS market for Intel compatible PCs

317 Possible explanations for the low price of Windows To hook consumers … but when will MS increase the price? Competition from installed base of Windows … but (i) very difficult to uninstall Windows (ii) consumers buy much better new PCs faster than traditional obsolescence rates would imply (iii) Windows price is small compared to the PC+Windows bundle To reduce pirating … but why is then MS-Office price high? Because it allows for higher prices of complementary goods …but MS does not monopolize all the complementary goods markets, therefore it would be optimal to charge the monopoly price on Windows Because of the existence of actual and potential competition

318 Monopoly may maximize social surplus when there are network externalities present under conditions of incompatibility; value of de facto standardization

Microsoft appealed, and was granted a stay of all parts of the District Court decisions until the appeal is heard Summary of court of appeals decision:Microsoft’s breakup and other remedies imposed by the District Court were vacatedMicrosoft was found liable of monopolization of the operating systems market for PCsMicrosoft was found not liable of bundlingMicrosoft was found not liable of attempting to monopolize the Internet browser marketThe district court judge Thomas Penfield Jackson was taken out of the case for improper behaviorThe case was remanded to the District Court for remedies determination for the monopolization charge 319

Bundling issue The Appeals Court instructed the District Court to examine the bundling of IE and Windows (if plaintiffs bring it up) under “a rule of reason” where the consumer benefits of bundling are balanced against the damage of anti-competitive actionsIn face of the Appeals Court decision, DOJ decided not to pursue the bundling issue and announces that it will not ask for a breakup of Microsoft 320

The Settlement: On November 6, 2001 the United States, the states of New York, Illinois, North Carolina, Kentucky, Michigan, Louisiana, Wisconsin, Maryland and Ohio, and Microsoft announced a settlementCalifornia, Connecticut, Iowa, Massachusetts, Minnesota, West Virginia, Florida, Kansas, Utah, and the District of Columbia pursued the suit further to a full remedies trial (started March 11, 2002) in front of U.S. District Judge Colleen Kollar-KotellyThese States proposed making the source code of Windows and IE public, “freezing Windows” so that additional functionality would be sold as an additional good, making all APIs public, and other severe remedies.On November 12, 2002, Judge Colleen Kollar-Kotelly imposed the final judgment that had only small differences from the original proposed settlement 321

Settlement terms A. Provisions seen as favorable to Microsoft 1. No breakup 2. Microsoft can expand functions of Windows 3. No general restrictions on bundling 4. No mandatory disclosure of source codeB. Provisions seen as favorable to the plaintiffs 1. Broad scope of definition of middleware products (including browser, e-mail clients, media players, instant messaging software, etc.) 322

Settlement terms (cont.) 2. Requirement to partially disclose middleware interfacesMicrosoft is required to provide software developers with the interfaces used by Microsoft's middleware to interoperate with the operating system3. Requirement to partially disclose server protocolsThe settlement imposes interoperability between Windows and non-Microsoft servers of the same level as between Windows and Microsoft servers. 323

Settlement terms (cont.) 4. Freedom to install middleware softwareComputer manufacturers and consumers will be free to substitute competing middleware software on Microsoft's operating system.5. Ban on retaliationMicrosoft prohibited from retaliating against computer manufacturers or software developers for supporting or developing certain competing software 324

Settlement terms (cont.) 6. Uniform pricing of Windows for same volume saleMicrosoft will be required to license its operating system to key computer manufacturers on uniform terms for five years. Microsoft will be allowed to provide quantity discounts.7. Ban on exclusive agreements; contract restrictionsMicrosoft will be prohibited from entering into agreements requiring the exclusive support or development of certain Microsoft software 325

Multi-year investigation of MS by the EU on interoperability between non-Microsoft servers and Windows clientsbundling of Windows Media Player with WindowsThe EU (i) Found Microsoft liable on both issues (ii) Imposed a $609 million fine (iii) Required MS to produce a version of Windows without WMP (called Windows-N) but without a requirement to sell it for less than Windows (iv) Required MS to make public and license at a low price the communications protocols between Windows clients and non-Windows serversLess than 2000 copies of Windows-N were sold 326

Second EU case on browser In December 2007, Opera (an Internet browser) brings the issue of Microsoft bundling IE with Windows to the EUIn January 2009, the EU issues a “statement of objections” to Microsoft, alleging a violation of Article 82 EC for tying IE to WindowsGiven the previous decision on WMP, liability of Microsoft seems certainMicrosoft offers to sell Windows 7 without any browser pre-installed (users would use FTP to download browser)Proposal rejected by the EU in favor of a “choice screen” approach 327

Choice screen Is imposed through Windows update on all users in the EU who are using IE as a default (on all Windows products)Will allow users choice among IE, Firefox, Safari, Opera, and ChromeOn December 16, 2009, the Commission accepted the final choice screen proposal, and the matter endedIn my opinion, the choice screen should have been made available to all, not just IE usersPresent solution gives incentives to “bribe” OEMs to choose a non-IE default browser, so the choice screen does not appearBenefits the non-open-source non-IE browsers (Chrome) 328

Investigations of GoogleFTC investigation, leaked two weeks before presidential election to create the impression that Obama was tough on antitrust EU investigationsBoth, mainly on placement in organic searchAggregatorsPlacement of search results of Google affiliatesAgreements with Android manufact. 329

US investigation fizzledFTC accepted proposed commitment Google proposed to clearly label the links of its affiliates(A joke!)EU Investigation continuing and recently got extended330

4/15/2015, EU sues GoogleSent Google a “Statement of Objections” Allegations:Google favors its own comparison shopping services (for airline tickets, hotels, etc.) by placing them on top of the first organic search pageVertical issue, may be hard for EU to win331

Allegations cont.Google demoted placement of “aggregators” such as Foundem332

EU launched Android investigation. Allegations:Google required that smartphone and tablet manufacturers exclusively pre-install Google applications on their devicesPrevented the success of rival apps by bundling its own apps and services on Android devices with other Google applicationsPrevented manufacturers who want to use Google apps from developing and marketing modified and potentially competing versions of Android on other devices333

Week 12: Bank Networks Formation and Systemic Risk 334

335 Network Structure and Collapse The network gets formed in anticipation of “good” and “bad” times Bad times eliminate nodes with prob. Pb As Pb increases, banks form ex ante more dense networks expecting more widespread collapses in bad times Pb increase is more costly to the each bank because it has to create a more dense network ex ante If Pb is underestimated, the prob. of collapse of any node of the formed network increases

Possibilities of Collapse (= Node Leaving the Network)Individual nodes can collapse both in good and bad times Collapse of Bi increases the probability of collapse of a connected node BjThere can be cascading collapsesFor the same network structure, cascading collapses increase as the probability of shocks increases 336

337 Network Structure It pays to have more connections because more connections make it more likely to find a trading partner (to diversify risk (Allen and Gale)) and therefore less likely for any node to collapse In “bad” times, there are fewer nodes to connect to, so less diversification of risk But fewer connections could restrict a cascading collapse to only a cluster of nodes by limiting the propagation of the financial distress (Battiston et at.) Higher connectedness could also degrade credit worthiness and increase financial distress

338 Public Policy to Reduce Impact of Collapses in “Bad” Times Network-specific recommendations Collect data on interconnected banks (and other institutions that have similar functions) and identify the network structure Analyze the banking sector as a network structure Assess the risk to a bank from further-away neighbors in the network Identify central nodes whose collapse can have very significant impact on the network

339 Systemic Risk as a Network Issue Our Model First, banks B1, B2, …, Bn form a network by checking the credit worthiness of each other at a cost and in anticipation of subsequent trading Second, the state of the world is revealed, Good = “g” or Bad = “b” Let Pg < Pb be the prob. of unacceptably high credit risk A bank (node) with unacceptably high credit risk collapses (or, equivalently, no one trades with it) Third, in each of M periods, each bank receives a positive or negative shock with prob. q If, in the same period, directly connected banks Bi and Bj receive opposite shocks they trade at low cost If banks Bi and Bj are not connected directly but only through other nodes they trade at a higher cost, and the intermediate nodes benefit too If bank Bi receives a positive or negative shock and cannot find anyone to trade with, it trades with the Fed at a high cost

340