1. Internet Economics John Chuang School of Information Management & Systems UC Berkeley chuang@sims.berkeley.edu

2. John Chuang2 The Big Picture Supply Price(s) Market Structure & Mechanisms Welfare (surplus) Demand Producer Surplus Consumer Surplus Social Surplus {

3. John Chuang3 Why Study Internet Economics?  Internet has interesting economic properties  Resource allocation -Rule-based vs. pricing-based  Market structures -Interconnections -Horizontal mergers and vertical integration -Bandwidth markets  Policymaking -Sustainable competition -Universal access

4. John Chuang4 Outline  Economic characteristics of the Net  Resource allocation and pricing  Interconnection and industrial organization

5. John Chuang5 Economic characteristics of the Net  Public vs. private good  Economies of scale  Economies of scope  Network externalities

6. John Chuang6 Public vs. Private Goods  Private good -depletable and excludable -e.g., toothpaste, automobile  Public good -non-depletable and non-excludable -e.g., national defense, clean air, lighthouses  What about roadways, information, and the Internet?

7. John Chuang7 Public vs. Private Goods  Roadways: -non-depletable (until congestion) and non-excludable  Information: -Encapsulated: depletable and excludable -Non-encapsulated: non-depletable, but is it excludable?  Internet: -non-depletable (until congestion), but is it excludable?

8. John Chuang8 Economies of scale  Average cost declines as output level increases  Internet exhibits strong economies of scale  High fixed cost -e.g., trenching cost, up-front capital investment  Low/zero marginal cost -of sending an additional packet

9. John Chuang9 Traditional Goods & Services  Q* is optimal firm output  Can support N firms if market size (Q TOT ) >= NQ* $ Q AC Q*Q TOT

10. John Chuang10 Infrastructure Goods & Services  High FC, low MC  declining AC curve (economies of scale)  Therefore it is socially optimal to have the entire market served by a single firm (“natural monopoly”) $ Q AC Q TOT

11. John Chuang11  A monopolist: -is a price-setter, not a price-taker -maximizes producer surplus (profit), not consumer surplus  Alternatives: public utility or regulated monopoly -e.g., AT&T historically treated as regulated natural monopoly -rate regulation -structural regulation

12. John Chuang12 Competition  In a perfect competition: -all firms are price-takers -P = MC in the long run -inefficient firms with high MC will exit market -long term profits = 0 -consumer and total surplus maximized

13. John Chuang13 Technological Change  Natural monopoly may not last forever  Technological change may result in new cost curve: same market may now be optimally served by multiple firms  e.g., long distance telephony and the breakup of AT&T in 1984 $ Q Q TOT

14. John Chuang14 Economies of Scope  Significant joint costs of production for multiple goods/services  Examples: -GM plants produce sedans, SUVs, and minivans, etc. -Amazon.com sells books, music, and lawn-mowers, etc. -Internet supports multiple traffic types previously carried over different networks (telephony, radio, CATV, …)

15. John Chuang15 Service Differentiation Best Effort SLA email voice

16. John Chuang16 Service Differentiation Best Effort SLA QoS Aware Internet email voice

17. John Chuang17 Different applications have different needs – let’s build a smart network that supports them all! Why QoS?

18. John Chuang18 Why QoS? Economist’s View 1.Network congestion  negative network externality 2.Differing willingness-to-pay (WTP)  quality differentiation allows price discrimination 3.Economies of scope cost savings (e.g., statistical multiplexing)

19. John Chuang19 Network Externalities  Externality: value (including costs and benefits) of a good/service not fully reflected in its price -e.g., the price of an automobile does not include the economic impact of its potential to pollute  Network externality: value of the network is a function of the network size

20. John Chuang20 Positive Network Externalities  Value of network increases with network size -e.g., telephones, fax machines, email clients -Metcalfe’s Law: the value of a network is proportional to the square of the number of users (N^2) -Reed’s Law: the value of network grows with the number of possible sub-groups that can be formed (2^N)

21. John Chuang21 Negative Network Externalities  Value of network decreases with network size -e.g., due to increased likelihood of network congestion -During network congestion, each data packet incurs a social cost to other packets (e.g., delay, packet-drop)

22. John Chuang22 Summary  The Internet as a public good (?)  High fixed cost, low marginal cost (strong economies of scale)  Significant joint costs (strong economies of scope)  Positive/negative network externalities (demand-side economies/diseconomies of scale)

23. John Chuang23 Outline  Economic characteristics of the Net  Resource allocation and pricing  Interconnection and industrial organization

24. John Chuang24 Resource Allocation Goals (Objective Functions)  Technical efficiency -Performance (latency, throughput) vs. cost -Survivability (availability, redundancy) vs. cost  Economic efficiency -Social surplus -Pareto efficiency  Other objectives -Profit (producer surplus) -Penetration/usage s.t. cost recovery (e.g., universal service) -Equity, stability, predictability, etc. not necessarily aligned

25. John Chuang25 Rule-Based Resource Allocation  Example: TCP Congestion Control -All hosts reduce transmission rate when there is congestion -Some TCP-unfriendly implementations ignore congestion signal 1Mb/s 0.5Mb/s

26. John Chuang26 The Role of Prices  Allocate resources to maximize economic efficiency  Serve as feedback signals -Help users make efficient consumption choices -Help provider make optimal capacity expansions

27. John Chuang27 Pricing Network Services  Criticism of flat-rate pricing -Tragedy-of-the-Commons  Usage-based pricing -Metering costs -Users prefer predictable bills  Marginal cost pricing -MC=0 most of the time  Congestion-based pricing -Packets bid for service -Too costly to implement  Back to flat-rate?

28. John Chuang28 QoS and Pricing  QoS Pricing -Multi-class network requires differential pricing scheme -Otherwise all users select best service class  How about use differential pricing to implement QoS itself? -Paris Metro Pricing

29. John Chuang29 Desirable Properties of Pricing Schemes  Service provider’s perspective -Encourage efficient resource usage (incentive compatibility) -Low cost (implementation, metering, accounting and billing) -Competitive prices -Cost recovery  User’s perspective -Fairness -Predictability (reproducibility) -Stability -Transparency (comprehensibility) -Controllability (Delgrossi and Ferrari 1999)

30. John Chuang30 Outline  Economic characteristics of the Net  Resource allocation and pricing  Interconnection and industrial organization

31. John Chuang31 Local Exchange Carrier (LEC) Router Dial-Up ISP INTERNET Backbone Provider 1 Router Customer Premise Tandem Switch Inter-exchange Carrier (IXC) Long- Distance Network Corporate LAN Firewall Analog Modem Content Provider Server Router Remote ISP Point of Presence xDSL Modem Cable Modem Packet Network Headend Cable Network Local Loop DNS Local Ingress Switch Exchange Point Router Internet Service ProvidersCustomer Premises Internet backbonesTelephone Network Local Egress Switch Backbone Provider 2 Source: M. Sirbu

32. John Chuang32 Industrial Organization  Horizontal merger  Vertical integration/disintegration  Determinants: -Technological efficiencies -Transactional efficiencies -Market imperfections

33. John Chuang33 Vertically Related Markets  Upstream/downstream relationship  Examples: -Detroit: steel v. automobile -Software: OS v. applications -Telephony: local v. long distance -Internet: physical transport v. access v. content/services

34. John Chuang34 Vertical Integration  Good: -economies of scope savings -internalize transaction costs -reduce prices & increase total welfare  Bad: -if one component is monopolistic -foreclose competition in other component

35. John Chuang35 Vertical Integration: Telephony  Telephony was vertically-integrated industry  AT&T (Ma Bell) offered end-to-end solution  Divestiture in 1984 -Local service (the seven baby bells) -Long distance service (AT&T) -Customer premise equipment (CPE)  Removes hidden subsidies between local service (monopoly) and long distance (competitive)

36. John Chuang36 Vertical Integration: Internet  Different vertical components of Internet [Lehr98]: -Local access transport (LAT): PacBell, TCI (AT&T) -Retail Internet access provision (ISP): AOL, @Home -Wide area transport (WAT): AT&T, MCI-WorldCom, Sprint, Qwest, Level3 -Backbone Internet service provision (BSP): UUNET, AT&T, BBN  Note: AT&T vertically integrated across all four components

37. John Chuang37 Downstream Goods/Services  Internet data centers  Content distribution networks  Application service providers  Certificate authorities  Billing and payment services  Content providers

38. John Chuang38 Unbundling the Local Loop  RBOCs (e.g., Pacific Bell) own the local loop infrastructure and offers local phone/DSL service  Telecom Act of 1996 requires RBOCs to unbundle services from local loop access  Motivation: allow competitive local exchange carriers (CLECs, e.g., Covad, Northpoint) to compete against the incumbents  Difficult to implement/enforce; not sustainable

39. John Chuang39 Unbundling the Cable Plant  TCI owns/operates cable infrastructure (LAT)  @Home offers broadband Internet access over cable (ISP)  TCI and @Home are now one integrated entity: AT&T Broadband  Enters AOL… -wants to offer retail ISP service over AT&T’s cable infrastructure, in competition with @Home service -demands unbundling and open access to cable plant  Who wins?

40. John Chuang40 Horizontal Merger  Proposition: Economies of scale  Example: Internet Backbone -MCI-WorldCom (1998) -WorldCom-Sprint (2000; abandoned)  Objection: concentration leads to market power -Larger network has less incentive to interconnect, or to maintain a high quality interconnection -Larger network has negotiation power over smaller networks

41. John Chuang41 Fiber System Route Miles Source: Kende 2000

42. John Chuang42 Horizontal Merger  Example 2: Local loop  Seven Baby Bells Merging -SBC + PacBell + Ameritech -Nynex + BellAtlantic -Bell South -US West  1996 Telecom Act: unbundling and open access -competition in local exchange (e.g., Covad, Northpoint and other CLEC’s )  Facilities-based competition -e.g., wireless, cable, satellite, …

43. John Chuang43 Network Interconnection  Network externalities motivate network operators to interconnect  Different types of interconnection: -Peering -Multilateral -Bilateral (or private) -Transit  Issue of settlement -Peer = settlement-free = sender-keep-all (SKA)

44. John Chuang44 Peering Source: Kende 2000

45. John Chuang45 Multilateral Peering Source: Kende 2000

46. John Chuang46 Bilateral/Private Peering Source: Kende 2000

47. John Chuang47 Transit Source: Kende 2000

48. John Chuang48 Hot Potato Routing Source: Kende 2000

49. John Chuang49 Free Riding Source: Kende 2000

50. John Chuang50 UUNET Peering Policy  Need to meet following requirements to peer with UUNET (January 2001):  Interconnection Requirements -Geographic scope (> 50% of UUNET scope) -Traffic exchange ratio (not exceed 1.5:1) -Backbone capacity (> 622Mbps) -Traffic volume (> 150Mbps per direction)  Operational Requirements -24x7 NOC, fully redundant network, implement “shortest-exit routing”, …

51. John Chuang51 Interconnection Issues  Peer or transit? -Size (market share) important  Why multilateral peering fails? -Tragedy-of-the-Commons  What about advanced services? -Inter-domain multicast, inter-domain QoS, content peering, …

52. John Chuang52 Markets  Bandwidth Markets -Bandwidth is perishable -Bandwidth as tradable commodity  Contract terms -What: Diameter of pipe (Mbps) -Where: city A to city B -When/how long -Other: quality metrics (drop rates, latency, …)

53. John Chuang53 Bandwidth Exchanges  Two basic functions -Facilitate financial transaction -Facilitate physical delivery of traded BW  Three types of exchanges -Sole seller of bandwidth (e.g., Enron, Williams) -Neutral facilitator of member trading (e.g., Band-X, RateXchange) -Member-managed exchange (e.g., Bandwidth Financial Corporation, Commerex) Source: Mindel and Sirbu 2001

54. John Chuang54 Example: NY-London DS3, US$/month, 1-year contract Source: RateXchange

55. John Chuang55 Commoditization Trend Lines CommodityTiming Crude Oil  OTC  Futures Market  Derivatives Late 1970’s 1983 1985 Natural Gas  OTC -Between Pipelines -Intermediaries  Futures Market  Derivatives Early 1970’s Mid 1980’s 1990 1991 Electricity  OTC -Between Utilities -Intermediaries  Futures Market  Derivatives Late 1960’s Early 1990’s Mid 1990’s Source: RateXchange

56. John Chuang56 Commoditization Trend Lines CommodityTiming Telecom  OTC -Between Utilities -Intermediaries  Futures Market  Derivatives Late 1980’s Mid 2000 TBD Source: RateXchange

57. John Chuang57 Other Markets?  Distributed processing (P2P) -SETI@Home, entropia, Popular Power  Distributed storage/caching  Distributed object services