Embrace, extend and extinguish
Chun-Hui Miao
University of South Carolina
Introduction
• A two-sided market
• Platform composed of components
– Consumers buy clients
– Content providers buy servers
– Purchases are uncoordinated
– Components can be from different producers
Embrace, extend and extinguish
•
Kerberos - an encryption technology to
connect users to servers
1. Early versions of Microsoft Windows use
standard Kerberos;
2. Kerberos on Windows 2000 is not fully
interoperable with non-Windows Servers;
3. Users can only log on if both PCs and servers
run Windows 2000.
•
Other examples involve Netscape, Java
Overview
• Questions
– Why use tying to “extinguish” complementors?
– Why a phased response?
• Results
– Because price squeeze is impossible if customers on
two sides make uncoordinated purchases
– When technology gap narrows, embrace, extend,
extinguish
– not necessarily lower welfare
Literature review
• Price squeeze
– Chicago school, Ordover, Sykes and Willig (1985),
Whinston (1990)
• Tying in two-sided markets
– Rochet and Tirole (2003), Choi (2006), Amelio and
Jullien (2007)
• Competition in two-sided markets
– Armstrong (2006), Doganoglu and Wright (2006),
Carrillo and Tan (2006)
A static model
• A monopolist (MS) in client - A
• Two servers available - B and B’
• A consumer and a content provider trade and share
the surplus
– d if B’
» 0 ≤ d ≤ 1 (compatibility)
– γ if B
» γ < 1 (technology gap)
– A consumer (content provider) gets 1-s (or s) share
Preferences
• Identical consumers
• Content providers uniformly distributed on [0,1]
– server producers at the opposite ends
» Same marginal cost, normalized to 0
» Fixed cost F
• The more interaction with the other side, the more
to gain
The game
t
γ
Compatibility (A, B’)
Price of B
Price of B’
Price of A
Markets clear
The monopolist’s problem
• x₁: market share of B
– Decreases with d
client
server
• Profit = x₁(1-s)γ + (1-x₁)(1-s)d + x₁Ps
= (1-s)d + x₁[(1-s)(γ-d) + Ps]
↑d
↓d
+ only if γ large
Optimal degree of compatibility
Profits
•
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
F
0.2
0.0
0.0
0.1
0.2
d0
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Degree of Compatibility
The result
Profits 2.0
Embrace
Extend
Extinguish
1.5
1.0
0.5
0.0
0.2
0.3
0.4
0.5
0.6
0.7
Server technology
Shortcomings
• A static model, but the strategy is inherently
dynamic
– Cannot use the same A to embrace and
extinguish
– If use A1 to embrace, A2 to extinguish, why
anyone wants A2?
• Also, what about multi-homing by content
providers?
Extension to a dynamic setting
• Period 1
– A1, first generation of client
– B1’, open standard, competitively supplied
• Period 2
– An improved A2
– B2 by MS, the same quality as B1’
Timeline
T=1
Compatibility (A1, B1’)
T=2
Compatibility (A2, B1’)
Prices of A2 and B2
Content providers buy B2?
Consumers buy A2?
Preferences
• Content providers
– When connected, s
– Can multi-home, use B2 while keeping B1’
• Consumers
– not upgrade, (1-s)
– upgrade, u + (1-s)
» The second term only available when connected
– u uniformly distributed on [0, a]
» Some consumers will upgrade even if they lose the ability
to connect
Mechanism
• Sell an upgrade A2 incompatible with B1
– A negative externality on content providers
when consumers upgrade
• Offer a low price to consumers
– guarantee that enough consumers will upgrade
– force content providers to upgrade
Results
• Iff s > 4/7 and a > 4/7, the monopolist sells both
A2 and B2, of which A2 is incompatible with B1
• A large s
– consumers have little incentive to keep the old version
– a big loss for content providers if they do not upgrade
• A large a
– More valuable upgrade
– Great heterogeneity among consumers so it is easy to
get a sufficient number to upgrade
Welfare
• Content providers are worse off
• But more consumers upgrade - mitigates
monopoly undersupply
Conclusion
• In two-sided markets, tying of
complementary goods can be profitable
• but welfare implication is ambiguous
The monopolist’s problem
client
server
• Profit = x₁(1-s)γ + (1-x₁)(1-s)d + x₁Ps
= (1-s)d + x₁[(1-s)(γ-d) + Ps]
• If s=0 (and t→0)
– Only γ if both goods are monopolized
– Let d = 1, Pclient = 1 (price squeeze)
» Profit = 1 > γ (one monopoly theorem)