Industrial Organization. Networks

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Industrial Organization. Networks"

Transcription

1 In the Name of God Sharif University of Technology Graduate School of Management and Economics Industrial Organization ( st term) Dr. S. Farshad Fatemi Networks

2 Network externalities What are network externalities? In what types of markets do they arise? How do network externalities affect the adoption of new technology by users? How do firms compete in markets with network effects? Will firms adopt compatible standards? Access problems in network industries How should access to a monopoly network be priced? Industrial Organization Dr. F. Fatemi Page 167

3 Introduction to network externalities Definition: a good is more valuable to a user the more other people also use it Vi= Vi(qi, n) 2 forms Direct: often communication-related Telephone, , computer software (file-swapping) Indirect or market-mediated Computer software (availability), systems, broadcasting Industrial Organization Dr. F. Fatemi Page 168

4 Issues Demand side: will consumers join the network (adopt technology)? Supply side: which technologies will producers choose and promote? Will standardization occur? Note: In this session, we focus on network from the point of adopting a new standard, the literature also contributed to social networks and attracted a lot of attention. Industrial Organization Dr. F. Fatemi Page 169

5 Model of a communications network Model: n users connected to a network Each user gains benefit α> 0 from forming a link with another user Each network user can form a link with every other user Utility of each user: u(n) = α(n 1); Utility increases with network size: u (n) > 0 Network of size N; private benefit of (N+1) th user = αn Industrial Organization Dr. F. Fatemi Page 170

6 Total value of network: V(n) = α(n 1)n As n, V(n) αn 2 : Metcalfe s law Marginal value of an additional user (to the whole network): ΔV(N) = V(N+1) V(N) = αn(n+1) α(n 1)N = 2αN Social benefit of an additional user > private benefit αn Rohlfs-Griffin factor = SB / PB = ΔV(N) / u(n+1) = 2 Industrial Organization Dr. F. Fatemi Page 171

7 Demand side: user coordination Assume technologies are competitively supplied at cost i.e. no strategic behaviour on supply side Demand side: coordination problem Consumers utility functions are interdependent: higher if both adopt In choosing whether to adopt, must anticipate what others will do (Coordination problem) Industrial Organization Dr. F. Fatemi Page 172

8 Consumers may have different preferences Expectations are important, and tend to be self-enforcing Consumer bases decision on expected size of network Expect no-one else to join value to each consumer is low (no-one joins) Look for a fulfilled-expectations equilibrium Industrial Organization Dr. F. Fatemi Page 173

9 Model of user coordination Assumptions 2 users, i = 1, 2 Can either keep old technology or adopt a new one Incompatible technologies: size of network is technologyspecific User s utility (net of adoption costs), where n = network size, 1 or 2 Old technology: u(n) New technology: v(n) Industrial Organization Dr. F. Fatemi Page 174

10 Positive network externalities u(2) > u(1) and v(2) > v(1) Coordination is preferable u(2) > v(1) and v(2) > u(1) Game: both users simultaneously choose whether to adopt the new technology Industrial Organization Dr. F. Fatemi Page 175

11 2 eq. in pure strategies Both stick to the old technology u(2) > v(1): no unilateral incentive to adopt Both adopt the new technology v(2) > u(1): no unilateral incentive to stick This is true regardless of relative benefits, u(2) and v(2) Possible inefficiencies Excess inertia: v(2) > u(2) but both consumers stick Excess momentum: u(2) > v(2) but both consumers adopt Problem arises because of simultaneous decisions Industrial Organization Dr. F. Fatemi Page 176

12 Sequential decisions At t=1: Consumer 1 decides whether to adopt At t=2: Consumer 2 decides whether to adopt Solve backwards At t=2, consumer 2 will choose whatever technology 1 has chosen (Because u(2) > v(1) and v(2) > u(1)) Anticipating this, at t=1 consumer 1 will choose whichever technology gives higher utility Adopt if and only if v(2) > u(2) Fine as long as technology preferences are identical Industrial Organization Dr. F. Fatemi Page 177

13 Farrell and Saloner (RAND 1985) Joseph Farrell and Garth Saloner; Standardization, Compatibility, and Innovation; The RAND Journal of Economics (1985) Model with conflicting technology preferences User has taste (type) θ [0, 1] Measures willingness to join other user in adopting new technology vθ(2) uθ(1) increases with θ Industrial Organization Dr. F. Fatemi Page 178

14 Further assumptions v1(1) > u1(2): user with θ= 1 prefers new technology even if other does not adopt v0(2) < u0(1): user with θ= 0 always prefers old technology So coordination is a problem only for intermediate values of θ Information structure Each player s θ is independently drawn from uniform distribution on [0, 1] Asymmetric information: each knows own θ but not other s Industrial Organization Dr. F. Fatemi Page 179

15 Timing: 2 time periods, 1 and 2 In each period, users simultaneously decide whether to adopt the new technology (if have not already done so) Adoption is irreversible 3 possible strategies (A) Never adopt, regardless of other s decision at t=1 (B) Do not adopt at t=1; adopt at t=2 iff other has adopted at t=1 (jump on the bandwagon) (C) Adopt at t=1 (possibly initiating bandwagon) (NB: Adopting at t=2 regardless of other s decision is dominated by adoption at t=1) Industrial Organization Dr. F. Fatemi Page 180

16 Equilibrium strategy depends on θ Low θ: (A) do not adopt [recall v0(2) < u0(1)] Intermediate θ: (B) adopt conditionally at t=2 High θ: (C) adopt at t=1[recall v1(1) > u1(2)] Critical values of θ θ*: user is indifferent between strategies (A) and (B) θ**: user is indifferent between (B) and (C) Industrial Organization Dr. F. Fatemi Page 181

17 Solve for critical values θ* and θ** Condition for θ*: uθ*(1) = vθ*(2) Reasoning Neither strategy (A) or (B) prescribes adoption at t=1 Difference occurs at t=2 when other has adopted at t=1: (A) Stick with old: utility is u(1) (B) Follow suit: utility is v(2) Indifference requires uθ*(1) = vθ*(2) Industrial Organization Dr. F. Fatemi Page 182

18 Condition for θ**: (1 θ*) vθ**(2) + θ* vθ**(1) = (1 θ**) vθ**(2) + θ** uθ**(2) LHS: Strategy (C): adopt at t=1 Other user follows iff it has θ> θ*, i.e. with prob 1 θ* Thus, with prob1 θ*, user gets v(2) and with prob θ*, user gets v(1) RHS: Strategy (B): adopt conditionally at t=2 Other users adopts at t=1 iff it has θ> θ**, i.e. with prob 1 θ** Thus, withprob1 θ**, user gets v(2) and with prob θ**, user gets u(2) Industrial Organization Dr. F. Fatemi Page 183

19 Excess inertia may arise Both users have θ just below θ** Both prefer coordination on new technology: v(2) > u(2) Each would jump on bandwagon: both use strategy (B) But neither adopts at t=1, so no bandwagon starts Early adoption is a public good Solutions Communication between users; contracts or agreements Subsidy to early adopters to get the bandwagon rolling Industrial Organization Dr. F. Fatemi Page 184

20 Model with arrival of new consumers over time Not all consumers have same opportunities at a given time Initially, only old technology is available Installed base incurs switching cost of adopting new technology At date T, new technology becomes available New consumers may adopt either old or new technology If adopt new technology, bear cost of incompatibility (for a while) Industrial Organization Dr. F. Fatemi Page 185

21 Adopters of new technology exert two externalities Forward externality: increases attraction of new technology to subsequent adopters (+ve) (Excess inertia as before) Backward externality: installed base using old technology lose network benefits and face stranding ( ve) (Excess momentum) May get excess inertia or excess momentum, depending on Size of installed base: history matters! How quickly network benefits of new technology are realised Relative superiority of new technology Industrial Organization Dr. F. Fatemi Page 186

22 Parameter space divides into 3 regions Unique SPE: no adoption e.g. if large installed base and slow to realise network benefits May occur even if new tech highly superior: excess inertia Unique SPE: adoption New consumers benefit from new technology If cost of stranding to installed base is high, may get excess momentum Multiple eq: either may occur (depending on expectations) Industrial Organization Dr. F. Fatemi Page 187

23 Supply side: strategic behaviour So far we have assumed that technologies are supplied at cost by a competitive market Now consider imperfectly-competitive sponsoring firms Non-cooperative equilibrium Compatibility choices Users expectations must be incorporated into the analysis Industrial Organization Dr. F. Fatemi Page 188

24 Katz and Shapiro Model Michael L. Katz and Carl Shapiro; Network Externalities, competition, and compatibility; AER 1985 Michael L. Katz and Carl Shapiro; Technology Adoption in the Presence of Network Externalities; JPE 1986 Consider competition between oligopolistic firms Industrial Organization Dr. F. Fatemi Page 189

25 Model: duopoly i = 1, 2 Products are incompatible But otherwise are identical (perfect substitutes) Identical, constant production cost c Consumers Have unit demands Heterogeneous in basic WTP for the product: type s Homogeneous in valuation of network externality: v(qi e ) for product i with expected network size qi e Industrial Organization Dr. F. Fatemi Page 190

26 Consumer of type s chooses product i which maximises s + v(qi e ) pi Network size-adjusted ( hedonic ) price: p i= pi v(qi e ) ) Perfect substitutes: consumers choose product with lowest hedonic price min p 1, p 2 Prices can differ only if network sizes differ Cournot competition: firm i chooses output qi assuming Output of rival, qj, is fixed Consumers expectations, qi e and qj e, are given Market clears at hedonic price : p = 1 q1 q2 Industrial Organization Dr. F. Fatemi Page 191

27 Nash equilibrium: Consumers expectations about network sizes are fulfilled: qi e = qi Possible outcomes (with incompatibility) Symmetric duopoly Asymmetric duopoly Natural monopoly Multiple equilibria Importance of consumer expectations in selecting eqm Firm may become large simply because consumers expect it to Role of reputation, advertising, announcements in influencing expectations Industrial Organization Dr. F. Fatemi Page 192

28 Compatibility Choices Compatibility: valuation of network benefit is v(q1 e + q2 e ) K&S solve for Nash eqm in outputs with compatibility Total output is higher Making products compatible Agreement without side payments: both must benefit Agreement with side payments: total benefit must be +ve Produce an adapter: at least one firm must benefit; other may not Compatibility incentives Smaller firm has greater incentive to be compatible than larger firm Industrial Organization Dr. F. Fatemi Page 193

29 Access problems in network industries Network industries Large fixed cost of network operation Duplication prohibitively expensive natural monopoly E.g. gas transportation; electricity transmission and distribution; telecoms local loop Other parts of industry potentially competitive E.g. gas supply; electricity generation & supply; long-distance telecoms But to operate in these parts, competitors must gain access to the monopoly network Industrial Organization Dr. F. Fatemi Page 194

30 Access regulation May need to regulate access terms and pricing Especially problematic when network owner is vertically integrated into competitive sectors (e.g. retail services) Industrial Organization Dr. F. Fatemi Page 195

31 Access pricing: 2 cases Suppose network has fixed cost F and marginal cost ca Vertical separation Regulate access price, a, at cost Ideally, set a = ca: but then must subsidise F Without subsidy firm must recoup F: set a = AC (or LRIC) Vertically integration Asymmetry between incumbent and rivals in competitive sector Incumbent faces marginal cost of access, ca Competitors face access price, a Suggests setting a = ca to level the playing field Industrial Organization Dr. F. Fatemi Page 196

32 Access pricing with vertical integration Network owner must recover fixed cost F F is recouped from mark-up on retail prices If access is priced at marginal cost ca, competition in retail market will eliminate this mark-up fixed cost F cannot be recovered Suggests that access price must be above marginal cost Undesirable to encourage inefficient entry Want to encourage entry by firms that are at least as efficient as the incumbent But not those whose costs in competitive sector are higher than the incumbent s Industrial Organization Dr. F. Fatemi Page 197

33 Efficient component pricing rule (ECPR) ECPR:Baumol (1983) & Willig (1979) Costs Prices Vertically integrated incumbent Retail market: supplies a single output Wholesale market: supplies network access to retail competitor(s) Access is rivalrous: if a unit of access is granted to competitor, there is one less unit available to the incumbent Network: fixed cost F, marginal cost ca Downstream (competitive) market: incumbent has marginal cost cd Retail price is regulated at p; margin p (ca+ cd) to cover F Access price a: how should this be set? Industrial Organization Dr. F. Fatemi Page 198

34 The efficient access price a should be equal to Direct marginal cost of providing access, ca Plus opportunity cost (foregone downstream profit) to incumbent of supplying the input to competitors, p (ca+ cd) Result: margin rule Access price a = ca+ p (ca+ cd) = p cd The efficient access price equals the retail price minus the marginal cost of the competitive activity Equivalently, incumbent s margin in retail market (p a) should equal its marginal cost in that sector, cd Basis of margin squeeze test Industrial Organization Dr. F. Fatemi Page 199

35 Advantages of ECPR Ensures productive efficiency Rival entry occurs if and only if rival is at least as efficient as the integrated firm (i.e. its downstream MC is no higher than c) E.g. entrant with low downstream cost cl< cd Total cost per unit = a + c L = p c d + c L < p E.g. entrant with high downstream cost ch> cd Total cost per unit = a + c H = p c d + c H > p Revenue neutrality Entry does not reduce network cost recovery Implementation is relatively straightforward Requires information about downstream costs, but not demand etc Industrial Organization Dr. F. Fatemi Page 200

36 Criticisms of ECPR Revenue neutrality: a monopolist s charter Network owner is compensated for any loss of retail profit resulting from entry: rule preserves any monopoly profit Simplicity results from strong assumptions Entrants may offer new / differentiated products Or may be able to bypass the network, probably with less efficient technologies Then displacement of incumbent s product in retail market may not be one-for-one Can extend rule for differentiated products: a = ca+ σ(p ca cd) where σ= displacement ratio: no of lost sales to incumbent for each unit of access supplied (typically < 1) Industrial Organization Dr. F. Fatemi Page 201

37 Multi-product access: Ramsey pricing Efficient fixed cost recovery across differentiated products Rivals supply products that are differentiated from that of incumbent firm How should fixed cost F be recovered from the various products, whether supplied by incumbent or competitors? Ramsey pricing Inverse elasticity rule: mark-up on each product is proportional to its inverse elasticity of demand Revenues exactly cover costs Implies different charges to different services using the same network Informational requirements are huge: demand elasticities Industrial Organization Dr. F. Fatemi Page 202