Transcript EC 170: Industrial Organization

Network Markets
Chapter 17: Network Markets
1
Introduction
• Some products are popular with individual consumers
precisely because each consumer places a value on others
using the same good
– A telephone is only valuable if others have one, too
– Each user of Microsoft Windows benefits from having lots of
other Windows users
• Users can run applications, e.g., Word on each other’s computers
• More applications are written for systems with many users
• Network Effects or network externalities reflect such
situations in which each consumer’s willingness to pay for a
product rises as more consumers buy it
• Strategic interaction in a market with network effects is
complicated
Chapter 17: Network Markets
2
Monopoly Provision of a Network Service
• An early model by Rohlfs (1974) illustrates many of the
issues that surround markets with network effects
– Imagine some service, say a cable network, where consumers
“hook” up to the system but the cost of providing them service after
that is effectively zero
• Provider is a monopolist charging a “hook up” fee but no other
payment
• The basic valuation of the product vi is uniformly distributed across
consumers from 0 to $100. Consumer willingness to pay is fvi where
f is the fraction of the consumer population that is served
• The ith’s consumer’s demand is:
qiD =
Chapter 17: Network Markets
0 if fvi < p
1 if fvi  p
3
Monopoly Provision of a Network 2
• Consider the marginal consumer with basic valuation v~  pf
~
• The firm will serve all consumers with valuations  v
• With 100 consumers, solving for the fraction f of the
market served we have:
f = 1 - v~ / 100= 1 – p/100f
• So, the inverse demand function is:
Chapter 17: Network Markets
p = 100f(1 – f)
4
Monopoly Provision of a Network 3
• The inverse demand curve has both upward and downward
sloping parts. This means that there are two possible
values for the fraction of the market served at any price p.
$/unit = p
25
$22.22
20
15
10
5
0
0 0.2 0.4
fL
0.6
0.8
1
f
fH
Chapter 17: Network Markets
5
Monopoly Provision of a Network 4
• The Rohlfs model makes clear many of the potential
problems that can arise in markets with network effects
• 1. The market may fail altogether
– Suppose the firm must set a fee over $30 perhaps to cover fixed costs
– Network will fail even though it is socially efficient
• When half the market is served, the customers hooking up have vi ‘s
that range from $50 to $100 or fvi values that range from $25 to $50
• Average value is then $37.50, well above $30
• But as p rises to $30, f falls and so does average willingness to pay
• There is no price at which sufficient numbers of consumers sign on
that yields an average willingness to pay of $30
Chapter 17: Network Markets
6
Monopoly Provision of a Network 5
• 2. There are multiple equilibria
– At p < $25, there is more than one equilibrium value of f
– At p = $22.22 both fL(p) = 1/3 and fH(p) = 2/3 are possible f values
– Lower fraction may be unstable (tipping)
• This group is comprised of consumer with top one-third of vi values
• The addition of one more consumer will raise willingness to pay
sufficiently that consumers with the next highest third of vi values
will be willing to pay and we will move to the fH equilibrium
• The loss of one consumer will lower the willingness to pay of that
same top one-third and demand will fall to zero at p = $22.22
Chapter 17: Network Markets
7
Monopoly Provision of a Network 6
– If the firm needs to serve more than one-third of
consumers at a price of $22.22, fL is called a
critical mass.
• Low or free introductory pricing
• Lease and guarantee that if critical mass is not
reached, refund given
• Target large consumers with internal networks first
Chapter 17: Network Markets
8
Networks, Complementary Services & Competition
• Rohlfs model is a monopoly model but has clear
insights for oligopoly setting
– Market may fail
– Competition will be fierce—a firm that fails to reach a
critical mass isn’t just smaller than its rival—it dies
– Multiple Equilibria are possible—Betamax versus VHS
or Blu-Ray versus AOD DVD format—either system
may win
– Winning system is not necessarily the best one
Chapter 17: Network Markets
9
Price Competition and Network Effects
• Sometimes firms have separate networks instead of
connecting consumers to the same network
– Banks/ATMS
– Facebook/Myspace
• Consider to firms at the ends of a Hotelling line
–
–
–
–
A at zero, B at 1
Consumer valuation is V + k sA – tx – pA for firm A
V + k sA – tx – pA for firm B
Where si is the fraction of consumers buying from firm i
Chapter 17: Network Markets
10
Price Competition and Network Effects 2
• Sometimes firms have separate networks instead of
connecting consumers to the same network
Banks ATMS
Facebook/Myspace
• Consider to firms at the ends of a Hotelling line
– A at zero, B at 1
– Consumer valuation is V + k sA – tx – pA for firm A
– V + k sA – tx – pA for firm B
– Where si is the fraction of consumers buying from firm I
– sA = xm and sA = 1- xm where xm is the location of the
marginal consumer
Chapter 17: Network Markets
11
Price Competition and Network Effects 3
• Assume t>k > 0
• The marginal consumer is defined by
V +k sA – txm – pA = V + k sB – t(1 – xm) – pB.
• Using rational expectations we have (sA-sB=1+2xm)
2xm (t  k )  t  k   pB  pA 
Chapter 17: Network Markets
12
Price Competition and Network Effects 4
• Which gives demand and profits
 1  pB  p A  
q x N  
N

 2 2t  k  
D
A
m
 1  pB  p A 
  p A , pB   p A  
N

2t  k  
2
A
 1  p A  pB  
q  1 x N   
N

2t  k  
2
D
B

m

 1  p A  p B 
  p A , pB   pB  
N

2t  k  
2
B
Chapter 17: Network Markets
13
Price Competition and Network Effects 5
• And best responses
pB  t  k
pA 
2
pA  t  k
pB 
2
• So in equilibrium pA = pB = t – k
• So stronger network effects (higher k) intensifies
price competition
Chapter 17: Network Markets
14
Systems and Standards Competition
• Competition between networks does not always lead
to one survivor
• Each network may have its own system
—Compatibility issues
—What is gained and lost when consumers cannot use their
brand of the product on other systems?
• Competition to be the Industry Standard
—Firms may compete to have their system adopted as the
industry standard
—What are the implications of standards competition?
Chapter 17: Network Markets
15
Both
Fear ofand
beingTechnical
incompatible
can
Competition
Compatibility
switching to
lead to the inferior Nash Equilibrium—
Both
The fight over compatibility can lead to poor technical the new
neither firm switches because it
choices overall
technology is
staying
thinks the other won’t switch
a superior
withTwo
the possible
old
problems: Excess Inertia and Excess Momentum
Nash
technology
Equilibrium
is a Nash
Firm 2
Equilibriu
m Excess Inertia
Old
New
Technology
Firm 1
Technology
Old
Technology
(5,4)
(2, 2)
New
Technology
(1, 5)
(6,7)
Chapter 17: Network Markets
16
Both
Both staying
switching
Technical
Compatibility
2
with the old Again,
fear of being incompatible can
to the
new
technology
is of
In the case
excess
inertia,
each
firmEquilibrium
wants to adopt the
same
lead
to the
inferior
Nash
technology
technology as its rival but, fearful that the rival won’t switch
to
a superior
the new technology, each wrongly stays with the old is a Nash
Nash
It is also possible that there is Excess Momentum andEquilibriu
each
Equilibrium
wrongly switches to the New Technology
m
Firm 2
Excess Momentum
Firm 1
Old
Technology
New
Technology
Old
Technology
(6,7)
(2, 2)
New
Technology
(1, 5)
(5,4)
Chapter 17: Network Markets
17
Technical Compatibility 3
• The Excess Inertia and Excess Momentum cases apply to
market settings where the network gains from compatibility
and “connectedness” are large
– both firms want to adopt a common technology
– Difficulty in agreeing which technology both should use
• Sometimes firms will not have a preference to make their
technology the common standard or not to have a common
technology at all
– Different technologies loses compatibility
– But different technologies differentiates each product and softens
price competition, e.g., PlayStation 3 vs. Wii vs. X-Box
Chapter 17: Network Markets
18
Firm 1 choosing
Firm 1 choosing
technology 1 Technical
and
STRATEGIES:
Compatibility
4
technology
2 and
• two
build
an early lead
Firm 2there
choosing
Assume
are
technologies,
Firmby
1’sestablishing
technology
1 and
Firm 2 choosing
2) convince
Firm 2’s technology
technology
1 aislarge
the2 installed base; and technology
the suppliers of complements to 2 is the
Nash
Equilibrium
In Battle
of the Sexes
firms
stillpreferred
agree thattechnology
there
should
be a common
Nash
Equilibrium
adopt
your
standard but
each wants
to be the standard
preferred
by Firm
1 its own technologypreferred
by Firm 2
Firm 2
Battle of the Sexes
Technology 1
Technology 2
Technology 1
(10,7)
(6,5)
Technology 2
(5,4)
(8,12)
Firm 1
Chapter 17: Network Markets
19
STRATEGIES similar to before.
build
a large installed
base of the 5
Technical
Compatibility
Firm 1onchoosing
preferred
technology
with
your
name
it;1 and
Again,
assume
there
are
two
technologies,
technology
Firmand
1 choosing
1 and
make
sure
that you 1have
linedtechnology
up
suppliers
of
technology
2, but
technology
is probably
better
technologycomplements
2 and
so that you are the
one2who
Firm
choosing
In Tweedledum
and to
Tweedledee,
firms
want to differentiate
adopt thatthe
technology
Firm
2 choosinggets
technology
2 wants
is theto be
their products by choosing different strategies but each
technology
1 is the
thesuperior technology 1 Nash Equilibrium
the one with
Nash Equilibrium
preferred by Firm 1
preferred by Firm 2
Firm 2
Tweedledum and
Tweedledee
Technology 1
Technology 2
Technology 1
(3,3)
(8,5)
Technology 2
(6,7)
(2,2)
Firm 1
Chapter 17: Network Markets
20
If Firm 1 chooses
Technical
Compatibility
6
technology
2 then
There
is
no
Nash
Equilibrium
If Firm
1
chooses
In Tweedledum and Tweedledee each firm wantsFirm
superior
2 wantsbyto
(in pure
mayproducts
technology
1 then
technology
but really
care strategies)—Firm
about differentiating2their
adopt technology 2,
choosing
different
technologies
change or update
Firm
2 wants
to frequently
toobrother)
In Pesky Little
Firm 1 is the
its technology
to dominant
lose its firm (big
use technology
1, Brother,
that wants to limit competition
from
Firm 2 (little brother) by
“little
brother”
as
well
adopting a different technology. Firm 2 always wants compatibility
Firm 2
Pesky Little
Brother
Technology 1
Technology 2
Technology 1
(12,4)
(16,2)
Technology 2
(15,2)
(10,5)
Firm 1
Chapter 17: Network Markets
21
Public Policy and Systems/Standards Competition
• Public policy in the presence of strong network externalities
is complicated
• Low introductory pricing and bundling of complements
may look like anticompetitive
practices but are really just necessary to survive
• Decreeing a common standard forces government
to choose the winning standard. Governments are
not necessarily good at picking winners
• Should governments try to coordinate technology choices
or, instead, “let a thousand flowers bloom”
Chapter 17: Network Markets
22
Empirical Application: Network Effects in
Software—The Case of Spreadsheets
• Computer software is probably among those products with
important network features, e.g., the more people that use
Excel or PowerPoint the more usable and valuable they are
to any one consumer
• Can we identify network features empirically?
• A relatively early attempt is Gandal’s (1994) investigation
of spreadsheet program pricing
Chapter 17: Network Markets
23
Empirical Application: Network Effects in
Spreadsheet Programs 2
• A spreadsheet is a long-established business planning tool
– Originally a pencil-and-paper operation with sheets organized into
many rows and columns that could be summed either vertically or
horizontally to trace the impact of individual factors
– Computerized versions began to appear in 1980
• By the mid-1980’s there were eight or more different
spreadsheet programs on the market
– The dominant product was Lotus 1-2-3
– Each product had different features, e.g.,
• Graphing
• Ability to link entries in one spreadsheet to those in another
• Lotus compatibility
Chapter 17: Network Markets
24
Empirical Application: Network Effects in
Spreadsheet Programs 3
• A hedonic regression is a model of price determination that
explains a product price as a result of its key features rather
than explicitly model supply and demand
• Gandal (1994) estimates an hedonic regression for
spreadsheet programs over the years 1986 to 1991
– Postulates key characteristics that should affect spreadsheet price
– Identifies which characteristics are network features
– Explicitly considers the role of time and technical progress so that a
spreadsheet price index may be constructed
Chapter 17: Network Markets
25
Empirical Application: Network Effects in
Spreadsheet Programs 4
· Basic Features
LMINRC = a measure of sheer computing power
LOTUS = a 1,0 variable equal to 1 if it has the Lotus brand
GRAPHS = a 1,0 variable equal to 1 if it has graphing ability
WINDOW = number of windows program handles simultaneously
LINKING = a 1,0 variable equal to 1 if it links spreadsheet entries
Network Features
LOCOMP = a 1,0 variable equal to 1 if program is Lotus compatible
EXTDAT = a 1,0 variable equal to 1 if it can import external data
LANCOM = a 1,0 variable equal to 1 if it can link to a local network
Time Dummies
one for each year to pick up the pure effect of time (technology
improvement) on spreadsheet prices
Chapter 17: Network Markets
26
Empirical Application: Network Effects in
Spreadsheet Programs 5
Variable:
LMINRC
LOTUS
GRAPHS
WINDOW
LINKING
LOCOMP
EXTDAT
LANCOM
CONSTANT
1987
1988
1989
1990
1991
Coefficient
0.11 ( 1.59)
0.56
0.46
0.17
0.21
0.72
0.55
0.21
3.76(12.31)
– 0.06
– 0.44
– 0.70
– 0.79
– 0.85
t-statistic
( 4.36)
( 3.51)
( 2.14)
( 1.91)
( 4.28)
( 4.05)
( 1.65)
(–0.38)
(–2.67)
(–4.20)
(–4.90)
(–5.30)
Chapter 17: Network Markets
Network features
raise value of
spreadsheet
program
Technical
progress lowers
spreadsheet
prices over time
27
Empirical Application: Network Effects in
Spreadsheet Programs 6
• Gandal’s Results
– Demonstrate importance of network features for spreadsheet
programs
– Allow construction of a spreadsheet price index that controls
for quality, i.e., that reflects the pure passage of time
– Since dependent variable is ln Price, Hedonic Price Index is:
pit  e
Year
Index
 t YEARt
Implied Spreadsheet Price Index (1986 = 1.00)
1986 1987 1988 1989 1990 1991
1.00
0.94 0.64
0.49
0.45
0.42
Chapter 17: Network Markets
28