Networks and Standardization

2. Networks and Standardization

3. Networks and Standardization • Computer Science, • Geography, • Labor Sciences, • Interdisciplinary Research Project at Frankfurt University Networks as a competitive advantage (1997-2000) • Information Systems • Law and Legal Sciences, • Political Sciences, • Sociology • Business Administration

4. Networks and Standardization • Prof. Dr. Wolfgang König • Johann Wolfgang Goethe-Universität • Institut für Wirtschaftsinformatik • Mertonstraße 17 , D-60054 Frankfurt am Main • Telefon: +49 69 798-28594 • Telefax: +49 69 798-28585 • koenig@wiwi.uni-frankfurt.de • http://www.wiwi.uni-frankfurt.de/IWI/

5. Agenda The theory: Networks as a competetive adavantage A model: A standardization framework Applications and empirical data: Electronic Business

6. The theory: Networks as a competetive adavantage Standardization problems: EDI etc. Network effect theory: Basics Reconsidering network effect theory

7. Standardization problems "The good thing about standards is that there are so many to choose from"(A. Tanenbaum).

8. Standardization and compatibility • Every interaction is based upon communication • The exchange of information necessarily requires both, sender and receiver of a message to use a mutual language or set of communications standards (e.g. TCP/IP, EDI...) • Compatibility – and thereby standardization – is then a question of strategic importance for competitiveness • This interdependence results in coordination problems because actors, such as firms, do not know in advance when which standards will be implemented by other firms, if at all

9. Standardization and compatibility • Standards play a prominent role in many systems that are characterized by interaction or interrelatedness. • In information systems (e.g. Intranets, supply chains) standards provide for compatibility and are a prerequisite for collaboration benefits. • More generally speaking, standards can constitute networks. • Standard: any technology or product (software, hardware) incorporating technological specifications that provide for or require compatibility: • Products are said to be compatible when two products "somehow go together“ [Gabel 1991, 1].

10. Standardization problems "Natural selection is a process of standardization"[Perry 1955, 271] • Prominent examples of standardization problems • Electronic Data Interchange (EDI) standards, XML/EDI [ data] • Directory services (X.500) [ data] • Videocassette systems (VHS vs. Beta vs. Video2000) • Nuclear power reactors (Light Water vs. Heavy Water vs. Gas Graphite) • Automobiles: Gas vs. Gasoline • other examples: DNA, language, railway gauges,

11. Standardization problems • While the (opportunity) costs of a lack of integration are hard to approximate, OAG estimates IT companies to spend at least 40% of their IT budgets on integration. • As a consequence, corporate information management is increasingly occupied with coordinating standardization decisions as a basis for information and communication infrastructures.

12. Network Effect Theory

13. Network effects • Need for compatibility implies network effects: • Network effects describe the positive correlation between the utility derived from a good and the number of its users • „demand-side economies of scale“ • direct: Telephone • indirect: Complementary goods and services • Examples: • Telephone, Inter-/Intra-/Extranets, Supply Chains • Why interdependencies? • Integration (automation, E2E or A2A, c-commerce) only works with multilaterally agreed upon standards

14. General findings of network effect theory • In many cases, the existence of network effects leads to Pareto-inferior results in markets. Discrepancy between private and social gains from network participation can imply Pareto-inefficient equilibria • Positive network effects imply multiple equilibria and the market will finally lock-in to a monopoly situation [Arthur 1989; Katz/Shapiro 1985; 1986]. • Tippy networks [Besen/Farrell 1994] • Free-riding network coordination costs [Kindleberger 1983]

15. General findings of network effect theory • Excess inertia (start-up problem) can occur due to installed base effects as no actor is willing to bear the overproportional risk of being the first adopter of a standard; thus, start-up problems prevent adoption even of superior products [Katz/Shapiro 1985; Farrell/Saloner 1985; 1986; Kindleberger 1983, Dybvig/Spatt 1983] • Excess momentum can occur, e.g. if a sponsoring firm uses low prices in early periods of diffusion to attract a critical mass of adopters or due to stranding fears [Farrell/Saloner 1986]

16. Basic Standardization Games • Basic games used in game theory useful to explain general strategic situations in network effect systems • Especially, coordination games are characterized by multiple Nash equilibria (i.e. “somewhat” diverging interests as in the battle of the sexes game [Feess 1997, 49-50; Holler/Illing 2000, 11-12])… • …while discoordination games do not have any Nash equilibria at all (i.e. a “real” social dilemma [Rieck 1993, 53-64; Feess 1997, 50-54]).

17. Basic Standardization Games • e.g. PAL vs. SECAM • incentive to standardize first • French and Germany both standardized first ;-) • e.g. left-hand and right-hand driving • ...given there is no legacy infrastructure e.g. pescy little brother

18. Implications of the findings • ICT markets fail: • Discrepancy between private and social gains from network participation can imply Pareto-inefficient equilibria • Different policies for ICT markets? • Propensity to monopolize highly problematic for traditional anti-trust policies • But there are problems to governmental intervention: • "narrow policy windows" (only small time frame for intervention) • "blind giants" (vendors know technologies much better, have better judgment) • "angry orphans" (possibility of stranding imposes social costs and disincentives)

19. Path dependencies: a different perspective • "A path-dependent sequence of economic changes is one of which important influences upon the eventual outcome can be exerted by temporally remote events, including happenings dominated by chance elements rather than systematic forces" [David 1985, 332]. • Imortant influence of early „insignificant“events on diffusion outcome • Explanation for non-ergodic technology diffusion proceses • Examples: VHS vs. Beta, QWERTY vs. Dvorak keyboards

20. User-side economies of scale Externality property implies coordination problem Multiple equilibria Does the market select the right technology? Excess inertia Excess momentum Lock-in But is the future path of network effect development known? And is the ultimate network size known? Summary network effects

21. But if network effects were exhaustible… …or if there are costs increasing faster than benefits …there can also be efficient solutions characterized by the coexistence of multiple standards but...

22. Reconsidering Network Effect Theory

23. Summary network effects • Traditional network effect theory(ies): • Valuable insights into basic properties of systems subject to network effects (lock-in, tipping, market failure) • Very "intuitive" • But limited explanatory power, e.g.: • ONLY monopolies? • Do ALL ICT markets fail? • NO decentralized coordination? • NO individuality? • NO solutions to unfavourable equilibria? • UNLIMITED growth of network effects? • Not applicable to "real" problems: • What to do in my Intranet? • How to overcome the problems identified?

24. Theoretical drawbacks • Traditional approaches concerning network effects offer great insights to general problems about diffusion of standards... • ...but they fail to explain the variety of diffusion courses in today’s dynamic ICT markets, for example • coexistence of different products despite strong network effects • individuality and specific interaction of potential adopters within their personal socio-economical environment, heterogeneity of preferences of network agents • influence of agent size (large vs. small players) • potential decentralized coordination of network efficiency • strong players in communication networks force other participants to use a certain solution

25. Empirical drawbacks • ambiguous empirical evidence: difficult to empirically prove market failure due to network effects • Liebowitz/Margolis show that many historic cases (QWRETY) might be biassed, showing basically the opposite, i.e. efficient market selection

26. Theoretical drawbacks • Network effects versus network externalities • not all network effects are externalities [Liebowitz/Margolis 1994, 1995a] • especially network ownership ("sponsored networks") can internalize network effects • indefinitely increasing network effects • network effects independent of communication partners • mostly homogeneous preferences • no consideration of cost of network size • no differentiation between direct and indirect network effects in models

27. Alternatives to a Neo-classical Theoryof Network Effects ? • Institutional Economics • shares our criticism • neglects the explicit modeling of any behavioral assumptions • game-theoretical approaches still strive for analytical solutions • Exception: evolutionary branch of Game Theory • network effects imply complex system dynamics • Agent-based computational economics

30. A Standardization Framework Standardization model Simulation results Policy implications

31. Research questions • What are determinants of standardization processes in information networks? • What are typical phenomena of technology diffusion in networks and how can they be explained? • What are possible equilibria in standardization processes? • What are diffusion patterns of standards and what are the determinants of their paths? • How can corporate and governmental standardization decisions be supported? • What are theoretical requirements for understanding systems subject to increasing returns?

32. The basic standardization model • Standardization benefits • Common use of IT standards simplifies transactions between actors or eases the exchange of information • direct savings resulting from decreased information costs • strategic benefits (avoiding media discontinuities eliminates errors and costs, just in time production etc.) • More and better information can be exchanged between partners. Since information provides the foundation for any decision, better information implies better decisions. • Economically, this is represented as an increase in the information value.

33. The basic standardization model • Standardization costs • Implementation costs: Hardware, software, switching, and introduction or training • Coordination costs: The interdependence between individual decisions to standardize occasioned by network externalities can yield coordination costs of agreeing with market partners on a single standard • Concretely, these include costs for time, personnel, data gathering and processing, and control and incentive systems

34. The basic standardization model • A communications network is a directed graph without isolated nodes • The nodes represent the communications partners (e.g. human, machine, firm), characterized by their ability to process, save and transfer information • The network edges represent the communications relationships

35. The basic standardization model • The nodes represent the costs of standardization (Ki) for the respective network actors i • The edges show the costs of their communications relations (cij) with their respective partners. These costs include costs of information exchange, as well as opportunity costs of sub-optimal decisions • Cost reductions can be realized only when both communicating nodes i and j have introduced the same or a compatible standard

36. The basic standardization model • Thus, the decision problem arises which nodes should be equipped with which IT standard. There is a tradeoff between the node-related costs of implementing a standard and the edge-related savings of information costs Benefits of standardization ==> Information costs savings cij Disadvantages of standardization ==> Standardization costs Ki

37. Centralized coordination • Centralized coordination: • A central decision making unit (e.g. the state or a parent firm) is credited with the aggregate results

38. Centralized coordination • Example • Bilateral standardization: net savings of 9 MU 1 K1=10 2 K2=20 9 30

39. Centralized coordination • Centralized coordination determines a first-best solution, but... • Data problem: Given asymmetric information distribution and opportunistic behavior, complete information cannot be retrieved from all nodes at acceptable costs considering that the nodes' reporting induces resource allocation • Complexity problem: Even solutions to simple centralized coordination problems are difficult • Implementation problem: Systems of control and incentives guaranteeing that a given, preordained solution will be implemented by all actors also involve costs

40. Decentralized coordination Decentralized coordination: Autonomous actors make individual standardization decisions If E [U (i)] > 0, then i standardizes

41. Decentralized coordination anticipatory decision function

42. Decentralized coordination • Example • no standardization: • agent 1 suffers net losses from standardization • redistribution designs for costs and benefits neccessary (and advantageous) 1 K1=10 2 K2=20 9 30

43. Standardization problems • Much more complex in • multi-agent • multi-standard • multi-period scenario with • individually different information sets • network topolgies • installed bases (legacy systems) • individually different communication relations • agents of different size

44. Network simulations

45. Network simulations Simulation pattern: • network is initialized assigning normally distributed random values for Kiq to all agents i and cij to all network edges (communications relations) <ij> based upon network density V (0<V1) • determination of centralized solution (linear program is formulated and solved using JAVA-packages drasys.or by DRA Systems () and lp.solve 2.0 • http://www.opsresearch.com • http://siesta.cs.wustl.edu/~javagrp/help/LinearProgramming.html) • agent decisions (decentralized) according to scenario (choice sequence, number of standards, agent size, network topology=

46. Network simulations • Basic Model: • information cost savings c (ND(cijµ, 2)) • standardization costs K (ND(Kijµ, 2)) • network density V • installed base B (C) = 1000 (C) = 200 (K) = var. (K) = 1000 n = 35 T = 35 V = 1,0 B = 0,0 Q = 1 • Analogous results with other distributions, network size

47. agregate decision quality time mean expected standardization costs Basic Finding: The Standardization Gap

48. Basic Finding: The Standardization Gap • generally, frequency of standardization c.p. increases with cij and decreases with Ki. • less willingness to standardize in decentrally coordinated networks • Network-wide savings attainable through centralized coordination determine critical value for the costs of coordination above which a centralized solution is no longer advantageous.

49. Single choice:an agent can decide once to implement a standard and after this decision he is tied to it (he can choose not to standardize until - if at all - he finds it advantageous at one time) Reversible choice: Sudden decline of aggregate decision quality at low standardization costs (Q = number of standards) Influence of choice sequence scenario

50. Five diffusion patterns 1 2 • no agent standardizes • mixed solution (some agents standardize, some not) • complete standardization with the same technology ("monopoly") • complete standardization with different technologies ("oligopoly") • "dynamic equilibrium", (some) agents change technologies in a stable and repeated rhythm) 3 4 5