Network interconnection and competition Patrick Rey IE Business School and Fundación Ramón Areces Madrid, 5 April 2011
Dec 19, 2015
Network interconnection and competition
Patrick Rey
IE Business School and Fundación Ramón ArecesMadrid, 5 April 2011
Patrick Rey Network interconnection and competition 2
Introduction
TopicsTopicsInterconnection (wholesale level)
oconnectivity (capacity, QoS agreements, net neutrality)oaccess prices (termination charges)
Competition (retail level)olinear/nonlinear tariffs (large/small consumers)oon-net pricing (club effects)ocaller pays/receiver pays principles
(consumers/websites)o[entry / investments]
Positive / welfare analysis
Patrick Rey Network interconnection and competition 3
Issue: whether to connect / compatibility or qualityIssue: whether to connect / compatibility or quality(Katz-Shapiro AER 85, Farrell-Saloner Rand 85)
Snowballing, club effects (small shocks / large impact)
Inertia, lock-inorole of installed bases oswitching costs?
Coordination problems opositive: coordination devices onormative (Pareto criterion; heterogeneity?)
“one-way” interconnectionbackward/forward compatibility, downgraded version (read-
only)
Connectivity strategies
Patrick Rey Network interconnection and competition 4
Large networks are less eager to maintain Large networks are less eager to maintain connectivityconnectivityTrade-off: degraded interconnection
ohas a negative direct effect (network is less attractive)ohas an indirect strategic effect (rival also less attractive)
Strategic effect can dominate is network is large enoughorefusing / degrading interconnection (WC/MCI/Sprint)odeveloping closed / proprietary standards (iTunes)
Large? Installed base of locked-in customers, coordination
Connectivity strategies
Patrick Rey Network interconnection and competition 5
Connectivity strategies
Illustration: the WorldCom – MCI mergerIllustration: the WorldCom – MCI mergerCrémer-Rey-Tirole JIE 2000
Pre-merger: 4 backbones (WorldCom, MCI, Sprint, GTE) oall backbones have similar installed bases (1/4, say)opre-merger, all backbones aim at perfect connectivity
Merger between 1 and 2 (installed bases ½,¼,¼) → three connectivity strategies: oaccommodation: maintain connectivity (1,1,1)oglobal degradation (½, ½ , ½ ): never attractiveotargeted degradation + limit on transit (¾ , ½ ,1):
attractive
Patrick Rey Network interconnection and competition 6
Interconnection and competition
One-way versus two-way accessOne-way versus two-way accessOne-way access
oupstream: incumbent controls a bottleneck (e.g., local loop)
odownstream: incumbent competes with rivals→ essential facility doctrine, access regulation
Two-way accessonetwork operators compete for subscribersobut need each other to complete calls→ competitive bottlenecks
Patrick Rey Network interconnection and competition 7
Interconnection and competition
Cooperation or competition?Cooperation or competition?Interoperability requires cooperation
ostandards, protocols (QoS)ointerconnection agreements
… between competitorso“cooperation” may prevail over “competition”olack of cooperation from incumbents may hurt new
entrants
→ impact of interconnection prices on retail competition
Patrick Rey Network interconnection and competition 8
Mobile to mobile termination
Competing bottlenecksCompeting bottlenecksWholesale: bilateral interconnection agreements
operfect connectivity (no club effect)oreciprocal termination charge (access price): a
Retail: price competitionolinear / non-linear (two-part) tariffs (prepaid/post-paid,
subsidy)oon-net pricing (friends and family)oprices for sending and/or for receiving (US/EU)
Patrick Rey Network interconnection and competition 9
Framework
Cost structureCost structure
Origination and termination costs: co , ct
Reciprocal access charge aoon-net cost: c = co + ct
ooff-net cost: c = co + a = c + m, where m = a – ct
Network 2Network 1
oc a
co + ct
C2C1
ct – a
Patrick Rey Network interconnection and competition 10
Linear retail prices Armstrong (EconJ 1998), Laffont-Rey-Tirole (Rand 1998)
The access charge as a “collusive” device The access charge as a “collusive” device
Access mark-up m = a – ct inflates the “perceived” marginal cost
average marginal cost for network i: ci = c + αj m
In (symmetric, imperfectly competitive) equilibrium
price based on perceived cost : p = pe(c + m/2)o If access prices are set unilaterally: double marginalizationo If bilateral negotiation of a reciprocal charge: collusive device
a high enough termination mark-up sustains monopoly prices
Note: no termination revenue or deficit if “balanced pattern”
o[even if asymmetric market shares: α1 α2 m = α2 α1 m …]
oCompare with “bill and keep”
Patrick Rey Network interconnection and competition 11
Non-linear (two-part) tariffs Laffont-Rey-Tirole (Rand 1998)
Two-part tariff Two-part tariff ((FFii,p,pii))
Termination mark-up still drives up usage prices
pi = ci = c + αj m
But competition for subscribers dissipates profitso“waterbed effect” (Paul Geroski for F2M termination)
olower fixed fees, handset subsidies, …
If total demand is elasticoaccess charge is “neutral” for profit→ termination charge no longer acts as a collusive deviceobut decreases consumer surplus / total welfare (wrong signal)
Patrick Rey Network interconnection and competition 1212
Demand sideDemand side
Full participation α1 + α2 = 1
Balanced calling patterno each user calls all users with equal probability
o demand for calls per subscriber is q(p), indirect utility v(p)
→ variable surplus is wi = v(pi) – Fi
Horizontal differentiation à la Hotelling (differentiation t)
→ operator i 's market share is αi = 1/2 + (wi - wj)/2t
Full participation / waterbed effect
Patrick Rey Network interconnection and competition 1313
Termination rate does not affect industry profitTermination rate does not affect industry profit
Usage price to average marginal cost: p = c + m/2
Termination revenue dissipated on fixed fee: F = f + t – m/2
Consider stealing a caller from network rival network
o retail profit: neutral since price p∗ covers average cost c + m/2o termination between own subscribers: saves & loses mq∗/2o termination from rival’s subscribers: brings mq∗/2
→ net gain of mq∗/2: the fixed-fee is reduced by this amount
Industry profit is independent of m: 2Π = t
Full participation / waterbed effect
Patrick Rey Network interconnection and competition 14
On-net pricing Laffont-Rey-Tirole (Rand 1998), Gand-King (EconL 2001)
Reintroduces club effectsReintroduces club effectsPerfect connectivity
But, for same prices, users favour larger networks
Similar to compatibility / connectivity analysisSimilar to compatibility / connectivity analysisFiercer competition among similar networks
Profits are maximal for a termination charge
… below cost
However, also exacerbates the risk of “tipping”
Patrick Rey Network interconnection and competition 15
Receiver pays regime Laffont-Marcus-Rey-Tirole (Rand 2003)
Two types of usersTwo types of usersMobile in the US: senders/callers vs receivers/callees
Internet: websites send traffic / consumers receive it
Two pricesTwo prices
price for receiving traffic pr, demand Dc(pr)
price for sending traffic ps, demand Dw(ps)
→ volume of traffic (“balanced pattern”) Dc(pr) x Dw(ps)
Patrick Rey Network interconnection and competition 16
Receiver pays regime Laffont-Marcus-Rey-Tirole (Rand 2003)
The off-net cost pricing principleThe off-net cost pricing principleFor network i
πi = (pis + pi
r – c) DiwDi
c (on-net)
+ (pir + a – ct) Dj
wDic (incoming off-net)
+ (pis –a – co) Di
wDjc (outgoing off-net)
= [pis – (co + a)] Di
wDc + [pir – (ct – a)] Di
cDw
Under perfect competition, prices reflect off-net costs
ps = co + a, pr = ct – a
Patrick Rey Network interconnection and competition 17
Receiver pays regime Laffont-Marcus-Rey-Tirole (Rand 2003)
Pretty robust principlePretty robust principleArbitrary number of networks
Mixed traffic patterns
Customer cost heterogeneity: cost ctk, co
k for group k
Quality of service: pr+ = ct+ – a, ps+ = co
+ + a
Welfare implications (Ramsey pricing)Welfare implications (Ramsey pricing)
+ = ( , = ) =' '
r r s s
s r r st o r r s s
S p S pp p c p c a p c a
D p D p
Patrick Rey Network interconnection and competition 18
Policy puzzle
Theory: when networks compete in non-linear tariffsTheory: when networks compete in non-linear tariffsProfits are not affected by the termination charge
oLRT Rand 1998oheterogeneous users (Dessein Rand 2003, Hahn IJIO 2004)
Or maximal for termination charges at or below costoGans-King EL 2003oelastic demand (Dessein Rand 2003, Schiff RNE 2008)oasymmetric networks Carter-Wright (RIO 2003)
PracticePractice
Operators favour termination rates above cost While regulators push for low rates
o EU recommendation: termination rates at LRIC by 2012o France: from 14,94/17,89 c€/mn in 2004 to 3,0 / 3,4 in January
2011
Patrick Rey Network interconnection and competition 19
Spanish Mobile Termination Charges
18,98
14,6253
11,7214
7,00
21,2153
17,7572
13,1303
10,08
14,3649
11,736410,4174
8,749,61
7,87
9,058,03
13,0523
0
5
10
15
20
25
jul-01 31/10/2002 02/10/2003 21/10/2004 29/09/2005 oct-06 abr-07 oct-07 abr-08 oct-08 Abril-09 -Sept-09
euro
cen
ts/m
inut
e
Movistar Vodafone Orange Yoigo
GLIDE PATH
Patrick Rey Network interconnection and competition 20
Revisiting MTM termination
Link between FTM and MTM rates Link between FTM and MTM rates Armstrong and Wright 2008 Arbitrage (“hedgehogs”) / regulatory policies FTM: partial waterbed effect / double marginalization (unilateral
setting)
Asymmetric competition and entry barriersAsymmetric competition and entry barriers
Cazalda-Valletti 2008, Hoernig 2007 (predation), Lopez-Rey 2009
Users’ expectationsUsers’ expectations
Lopez-Hurkens 2010
Here: demand heterogeneity (heavy / light users)Here: demand heterogeneity (heavy / light users)call imbalance and different participation elasticities
Patrick Rey Network interconnection and competition 2121
Main insightsJullien-Sand Zantman-Rey (2010)
Assume that consumers exhibiting traffic deficit Assume that consumers exhibiting traffic deficit … have also a more elastic participation … have also a more elastic participation large users churn but always participate small users mostly receive calls and may not participate
pre-pay/post-pay, large/small users in post-pay, teenagers...
Then Then profits are maximal for MTM termination rates above
cost welfare is maximal for MTM termination rate
o also above cost o but below the level that maximizes profits
Patrick Rey Network interconnection and competition 2222
Light usersLight users Do not call but receive calls Elastic (symmetric) subscription demand βi = β(Pi,Pj)
o aggregate demand βTT(P) = 2βii(P,P) is decreasing in P o replacement ratio γ(P) = ∂β2(P,P)/∂P1 (╱ -∂β1(P,P)/∂P1) [0,1[∈
Heavy users Heavy users Total participation of 1 + βT
→ heavy users’ surplus becomes wi = (1+ βT)v(pi) – Fi
(they benefit from being able to call more light users
Full participation → operator i 's market share is αi = 1/2 + (wi - wj)/2t
Heavy and light users
Patrick Rey Network interconnection and competition 2323
Symmetric equilibriumSymmetric equilibrium
Network i 's profit is equal to
Πi = αi [(1+βT)(pi - c)q(pi) – (αJ+ βj)mq(pi) + Fi – f] + (αi+βi)αjmq(pj) + βi(Pi - φ)
The usage price maximizes the surplus from trade (LRT)
p1 = p2= p∗ = c + m/2; q∗ = q(p∗)
The equilibrium subscription fees are then
F∗ = f + t + (βT - 1)mq∗/2
Equilibrium: heavy users
Patrick Rey Network interconnection and competition 2424
Consider stealing a caller from network rival networkConsider stealing a caller from network rival network
No light user (βT = 0)
o net gain of mq∗/2 o the fixed-fee is reduced by this amount
With light users (β1 = β2 = βT/2 > 0), the reduction is smaller
o loses termination revenue from own light users: (βT/2)mq∗
o the fixed fee is augmented by this amount
Intuition
Patrick Rey Network interconnection and competition 2525
Equilibrium: light users
Network i's profit then writes asNetwork i's profit then writes as
Π i = (1/2)[F∗ - f + (1/2 + βi)mq∗ + (βi - βj)(m/2)q∗] + βi(Pi - φ)
Optimizing with respect to POptimizing with respect to Pii amounts to maximize amounts to maximize
G(Pi,Pj) = (Pi - C) βi(P1,P2) – C∗ βj(P1,P2)
C = φ - 3mq∗/4 represents the direct opportunity cost of attracting light users, net of the benefits from calls received on-net (half the charge) and off-net (full charge)
C∗ = mq∗/4 represents the opportunity cost of the rival’s light users corresponding to the termination deficit (the price charged for calling these light users covering only half of the termination charge)
Patrick Rey Network interconnection and competition 2626
Equilibrium: light users
Equilibrium subscription for light usersEquilibrium subscription for light users
where ε(P) denotes the own price elasticity of light users' subscription demand
*
* *
* *
3
4 1
PP mq
P P
Patrick Rey Network interconnection and competition 2727
Impact of termination rates
Increasing the termination rate has two effectsIncreasing the termination rate has two effects intensifies competition for light users via two channels
o waterbed : light users generate larger net termination revenue
o opportunity cost : preventing the other network from attracting a light user saves on termination costs
reduces competition on the heavy usersthe cost of losing a heavy user to competitor is weakened by the larger termination revenue that the calls to light users will generate (again because there is no equivalent volume of calls from light users to this heavy user)
Patrick Rey Network interconnection and competition 2828
Equilibrium profit
The operators’ profit isThe operators’ profit is
2Π∗ = t + (P∗ - φ + mq∗) βT(P∗)
PP∗∗ depends on depends on mm only through access revenue only through access revenue r = r = mq∗mq∗
→ the operators’ profit depends only on the access revenue r and is maximal when the termination markup is at monopoly level
mR = argmaxm mq(c+m/2)
Patrick Rey Network interconnection and competition 2929
Equilibrium surplus
Light users' surplus is increasing in access revenue Light users' surplus is increasing in access revenue rr
Effect on heavy users is ambiguous Effect on heavy users is ambiguous
heavy users' net variable surplus is SH = (1+ βT)v(p∗) – f – t – (βT - 1)mq∗/2
→ increasing the termination rate
raises network externalities from light users (through r = mq∗ and βT)
but also raises usage prices: p∗ = c + m/2
→ for m small, increasing m raises heavy users' surplus if light users' subscription demand is very elastic or if heavy users' usage surplus is not very elastic
→ → total welfare is maximal for a termination markup that is positive total welfare is maximal for a termination markup that is positive but smaller than the monopoly levelbut smaller than the monopoly level
Patrick Rey Network interconnection and competition 3030
Extension: on-net/off-net
With on-net/off-net price discrimination the analysis is With on-net/off-net price discrimination the analysis is similar with additional effectssimilar with additional effects usage price = marginal cost (on-net and off-net) revenue from light users' received calls is smaller (no gain
from on-net calls) tariff-mediated network effects intensify competition on heavy
users additional effect of light users on tariff-mediated network
effects
Welfare is maximal for a positive termination mark-upWelfare is maximal for a positive termination mark-up
Profit is maximal for a termination mark-up that may Profit is maximal for a termination mark-up that may be smaller or even negativebe smaller or even negative
Patrick Rey Network interconnection and competition 3131
Extension: FTM termination rate
Each subscriber generates an extra revenue Each subscriber generates an extra revenue from FTM terminationfrom FTM termination
waterbed effect on both userso both subscription fees decreaseo reduction can be larger for either type of user
the presence of light users
… limits the waterbed effect on heavy users
industry profit increases
Patrick Rey Network interconnection and competition 3232
Conclusion
The level of termination rate has contrasted effects on the The level of termination rate has contrasted effects on the intensity of competition for different types of consumersintensity of competition for different types of consumers high termination rates benefit small users due to more
competition than even the waterbed effect suggests
high termination rates reduce the intensity of competition for large users
Consistent theory explaining a positive effect of TR on profitConsistent theory explaining a positive effect of TR on profit
Regulation is useful but the rate should be set above costRegulation is useful but the rate should be set above cost
How much above costs? → need for empirical workHow much above costs? → need for empirical work
Effect of Receiver Pays Principle? Effect of Receiver Pays Principle?