,%SXCH»^^_^^
3
"&
UbkAiilES,
V
^
JS
^ Or TBC^
3\ll
i'T'.''.
"""C*
.1AR3 01990
Management m
the 1990s
Economic Issues
in
Standardization
Joseph Farrell
Garth Saloner
90s: 86-023
October, 1985
Sloan WP #1795-86
Department of Economics WP #393
®
Massachusetts Institute of Technology
Management in the 1990s
Sloan School of Management
Massachusetts Institute of Technology
J
Management
in the 1990s is an industry and governmental agency supported
research program. Its aim is to develop a better understanding of the managerial
issues of the 1990s and how to deal most effectively with them, particularly as these
issues revolve around anticipated advances in Information Technology.
Assisting the
work of the Sloan School
working partners
in
scholars with financial support
and
as
research are:
American Express Travel Related Services Company
Arthur Young and Company
British Petroleum Company, p. I.e.
BellSouth Corporation
Equipment Corporation
Eastman Kodak Company
General Motors Corporation
International Computers, Ltd.
MCI Communications Corporation
United States Internal Revenue Service
Digital
The conclusions or opinions expressed in this paper are those of the author(s) and
do not necessarily reflect the opinion of Massachussetts Institute of Technology,
Management
in
the 1990s Research Program, or
its
sponsoring organizations.
ECONOMIC ISSUES IN STANDARDIZATION
Joseph Farrell*
and
Garth Saloner**
By
Number 393
•GTE Labcratcries
,
_no., ^0 Sylvar. Ri
October 1985
.
,
Walthan MA
0225^.
••Massachusetts Institute of Technology. Saloner gratefully acknowledges
research support froc tne National Science Foundation (grant IST-8510162)
1.
Introduction
The subject of compatibility and the economics of standards has become
very topical in the telecommunications industry since the AT&T divestiture and
with the groving importance of international traffic.
tion
of such
new
services as
standardization activity,
deal of
and video
ISDN
and offers
Moreover, the introduc-
communications
demands a
opportunities for
great
good and bad
decisions on standards^.
Standardization and its problems are not unique to telecommunications.
Compatibility may be important even if there is no physical network. For example,
one can use nuts and bolts from different manufacturers;
same records on different brands of gramophone,
to list
and so forth.
one can play the
It is also easy
situations in which more compatibility would be useful.
travelers
find
incompatibilities
weights and measures,
and in
lenses from Nikon, and
computer is of little
a
m
and
Buyers
currency.
in
electrical
of Pentax cameras
outlets,
in
cannot buy
library of software accumulated for an IBM personal
value if its
Other examples are cited
language
International
m
ov.-ner
Berg (1955),
decides to buy an Apple computer^.
Farrell and Saloner
(19£5a,b),
Katz
and Shapiro (19o5a,b).
scale
tions,
Such benefits
from
doing
demand
side.
This
on
the
for example:
the
same as others do create economies of
raises
a
number
of
important economic
ques-
2
Does standardization retard the adoption
1.1
may
One
reasonably
externalities")
("network
early adopter of new,
fear
when
that,
important,
are
incompatible
no
new technology?
of superior
benefits
tlie
will
user
technology,
since
lose the benefits of market compatibility. Tiius,
be
to
cor.patibility
willing
will be
it
to
be
an
reluctant
to
an industry could get trapped
in an inferior technology.
A related issue
m
(R&D)
to
is
order to develop
adept
a
the inceritives to undertake
researcli and development
better new technology.
the market may
a
superior in\-ention,
to develop inventions will be
When should
1.2
desirable
removes
courages
also
choice
1.3
late,
open up markets
that
may
m.ay
Who
if
itself
for
cur
there
r.mds
the
standard be
the
incentives
v.-ho
are
are else
about
early
set.
is
Early standardization
and thus
dov.'n,
en-
Standardization of interfaces can
unable
or
reasons to
the
standardization
optimal
unwilling
v.-ait.
to
compete
with
Inf crmstic. will flow
standard
to
set.
An
early
be hard to change later.
sets the
standards?
In some industries,
standards are set by formal industry groups accord-
ing to well-specified procedures. For e.xample,
and the
that
standard to settle
technclogy.
those
for
desirable,
is
the same
wait
to
Hcvever,
change
fear
inadequate.
early adoption of the
whole syster.E.
m
than
incentive
the
to
fail
standard be set?
standardization
Since
more
a
there are grounds
Since
CCITT have
this
rcie.
(Of
course,
m
telecommunications, the EZSA
different market participants have
3
different amounts of power in the process).
leader
sets
facto
de
standards.
For
In
example,
other industries,
it
that IBM has standard-setting pov;er in the computer
is
an industry
generally acknowledged
industry. In other cases,
buyers get together to demand standardization and even specify standards:
Hemenway (1975), especially chapter
1.4
see
2.
Can standardization have anticompetitive effects?
(a)
Advantage
Size as an
Because
size
confers an advantage
in
standardization,
large firms may
have various strategies available that will manipulate the market to their ad-
vantage and to their rivals'
product preannounce-
For example,
disadvantage.
ments and standard-sv;itching can become competitive tools.
There are
some grounds
for concern
(see FTC,
pecially design (as distiguished from performance)
19S3) that
standards, es-
may be
quality standards,
used to prevent or discourage entry, particularly entry of new technology.
(b)
Essential Facility Questions
VJhat
should antitrust policy say about compulsory admission of competi-
tors to a network (multiple listing service
ervations
systems in
the
airline market)?
produce advantages from market share itself,
dominance. Exclusion from
ket.
Likev.'ise
delay,
or
,
v;e
a
m
the real estate market,
Because
or res-
network externalities
may
they may create or extend market
network may in practice be exclusion from the mar-
can ask v;hether dominant
firms have
an
incentive
to
bias,
sabotage the process of cooperative voluntary standard setting,-
whether de facto standards tend to erode, extend or preserve monopoly power.
or
4
this
In
survey
we
paper,
first part of question 4. Question
search (Farreli and Saloner,
The paper
2
is orgariized
technology.
petition.
In
In
on
the
Section
Section
5,
is
as
the
the subject of some of our current
re-
follows.
In
Section
Section
In
we discuss the
we
and
on
market's willingness
4,
question
on
1
1965e).
efits and costs of standardization.
standardization
recent work
some
to
3,
we
2,
we
discuss some
focus on the
switch
from an
benefits and standardization. Section
a
6
effects of
old
to
a
new
effects on market structure and com-
consider the incentives to undertake
development (R&Z) of new technology in
bei:-
research and
market characterized by com.patibility
is
a
brief conclusion.
2.
Benefits and Costs of Standardization
We will describe three types of benefit from standardization: cost sav-
competitive
ings,
effects,
network
and
externalities.
will
We
then mention
some social costs that are less obvious than the benefits.
2.1
Benefits of Standardization
2.7.7 Cost Savings
By allowing
greater scale
to exploit economies of scale
use
is a
of interchangeable parts,
economies
(enabling
different manufacturers
in using a common supplier),
and by allowing the
standardization reduces production costs.
staple of economic history; see for instance Landes (1969,
This
1983), Hemen-
way (1975)^
Standardization also saves on the costs of learning how to use
For
instance.
Brock
(1975)
a
good.
discusses how pressure for standardization of the
programming language COBOL increased when machine time became cheaper relative
to programmer time,
making it less important to design programs that run effi-
ciently and more important to make it easier to
v>;rite
or transfer programs.
2.7.2 Competitive Effects
A
matter
corrjnon vie\:
among economists'* is that "plug-compatibility"^
for competition,
since
there will
be
competition in
overall systems level. We do nor take that view.
to entry
For
at
example,
the
the
market
for
ski
case
at
the
There will often be barriers
systems level that v;ould not exist
consider
any
does not
bindings
there might have been competition among incompatible
at
and
the
ski
components level.
boots.
Although
systems ex ante,
incom-
patibility destroys competition ex post'. Thus, we should recognize that there
may be effects of standardization on competition:
6
When the products of different makers are standardized,
•
m
directly
price and performance. Lack of standardization is
uct differentiation.
If variety is mostly "spurious",
erse tastes but merely splits
tlie
market,
a
if it represents the
different approaciies
it
is
to
firms
ha\'e
compete
whether it is better to
spare
Ofren,
or
performance
them compete
o)i
of
advantage
m
this
industry,
cars dc
thai'i
th.e
m
an
"af termarket"
peripheral devices.
or
replacement fenders for
crigmai manufacturer (car
m.aker)
tary input),
it
possible
source of after-sale profits;
is
a
is
notorious that spare parts have
car
has an
a
indeed,
m
much higher prcfit
Sometimes such pricing policies are inefficient (for in-
.
that
is
anticipated by buyers,
tc
ccr.r.it
it
may be prcfitable for
to low or reascnahle prices for
Standarcizaticn will achle^•e this'.
afterparts.
•
and if
(and r.utually beneficial)
the seller
As
well
as
protecting against
exploitation
may go bankrupt, or suffer
tiieir
purchases'.
See
a
a
supplier
crippling strike, buyers will v;orry about support
for
instance
auto parts industry before the ASME
this prctlem,,
ronopcly,
from e>:-post
standardization protects against orphaning. If it is perceived that
avoids
a
because buyers will inefficiently substitute away from the complemen-
stance,
for
or
taking the market (because he makes the original fenders). Absent
Standardization,
the auto
zr.c
better
product design.
such item.s (for instance,
natural monopoly,
a
is
standard product,
a
complementary inputs,
replacemer.t parts,
the m.arket for
model) IS
margin
liave
and
trying out of
not clear whether it
Standardization can commit producers to compete
•
for
price
on
then
of prod-
then this differentiation is ineffior
problem,
forir,
does not serve div-
i.e.
cient. If variety is valuable in itself,
to a
they compete
and thus
Hemenway's
discussion
(1975)
achieved standardization.
enhances competition,
since
a
of
the
Standardization
seller no longer
7
need be seen as both financially secure and committed to the industry in order
to
sell
against these
themselves
buyers
Sometimes
product.
a
problems.
insist
This
on
"second-sourcing"
equivalent
is
to
protect
guaranteeing
to
(lim-
standardization of their selected technology. For instance, in choosing
ited)
Intel to supply an essential component of the IBM PC,
licence its technology to
•
a
second supplier'.
Standardization can help in, or help replace, regulation.
telecommunications have been
tance
Conceivably,
carriers.
some
aspects
Long-dis-
deregulated in the United States
(partly)
by requiring local telephone companies to interconnect
tance
IBM insisted that Intel
of
local
\.'ith
non-AT&T long-dis-
telephone
service
could
also be deregulated if sv;itching protocols v;ere standardized and suitable pro-
vision
v;ere
exchange.
made for interconnections betv;een rival part-networks in the local
The
traditional view of telephone service as
a
natural monopoly is
based on the inefficiency of having duplicate netv;orks. This assumes that competing telephone companies would not have interconnection
bility).
It
is
While
possible
this may be
that
a
requiring
(a
form of compati-
plausible result of unregulated competition^".
interconnection would make
it
unnecessary
to
regulate other aspects of competition.
2-1.3 Network Externalities
Under this heading we group together
make
a
a
number of different effects that
large "network" of other compatible users desirable.
•
A product may simply be more valuable to each buyer the more others
have the product.
Telephojies
,
of course,
are an example.
Language is another
example: at is more valuable to speak English,
across who speak
franca
the
of
English,
nov see English becoming
and indeed we
where French and before
world,
the more other people one corres
that
the
linaua
Latin used to be used for
international discourse.
It
•
worth learning
IS more
people have the product.
to
type
on
the
Dvorak
a
(It is worth fev people's while,
keyboard,
A ccT.plementary
people have
as more
of software
m.ucl'.
cf
the
computers. The
anoti"ier
absolute
efficiency
the netv;ork externalities in-
An example of
ir.cer.tive
example.
the
m.ore
Tnis netv.-ork externality was
com.pany
willing to buy
rr.arket
v.-ith
m
tiie
late
ccrpetition
m.cre
from,
a
seller v:ith
other sellers. Tnere is thus
their products compatible
••
.
Katz and Shapiro
decision on compatibility when there are costs of
making products compatible. Eerc (19££) discusses
ity m.akes
a
1577.
for sellers tc make
discuss
the provision
importance of ready availability of
large m.arket share or compatibility
a
this is
success of the Singer sewmc machine
All these beriefits make buyers
(19£5a)
to learn
product may be more readily or more cheaply available
product is
e
1670s. See Chancier,
ar.
due to
is
the origir.al product.
for personal
repair net^'crk for
either
apparently,
)
•
behind
the more
a
considerable
despite
gams: David (19£5a) argues that this
volved.
product,
skill associated \;ith
direct, and firms
a
m.odel
m
trade cff the
com.patibility against this increase in competition.
v.-hich
gam
m
compatibilvalue from
2.2 Social Costs of Standardization.
The costs
of standardization
below the possible
discuss
are
from
loss
less
obvious than
reduction
in variety.
the
benefits.
There
possible loss from reduced innovation, which we discuss in Section
When tastes differ,
ferent tastes
ization.
v;hen
a
tradeoff.
a
served; but by hypothesis there
to be
There is
standardization involves
are benefits
is
We
also
a
3.
We want
dif-
to standard-
considerable literature on the economics of this tradeoff
the benefit to standardization is production economies of scale ^^.
has been little work on the corresponding problem
ization for other reasons,
we care about standard-
v.-hen
economies of scale associ-
the demand-side
such as
There
ated with network externalities.
In
a
sim.ple
model
we
(19S5c)
that there are two types of users,
can
ask the
following question.
Suppose
group A and group B, each of whom prefers
a
different standard. Consider three possible outcomes: both standards could be
provided (thus catering to diverse tastes),
or only
group A's preferred stan-
dard could be provided, or only group B's.^^ We show that there is a tendency
for too little standardization,
optimum is for
a
two incompatible
reverse cannot
single standard to exist,
it can happen that the
but the unique equilibrium is for
standards to persist (one for each type of buyer),
happen.
buyers on the other is
patible.
in the follov,-ing sense:
However,
Intuitively,
alv.-sys
there
the
externality
im.posed by
one
while the
group
of
negative if the first group chooses to be incom-
can be
multiple equilibria,
and it is
possible for
compatibility to be an equilibrium even though the optimum is to have incom-
patibility and variety.
Perhaps in practice this is most likely to be
a
prob-
lem v;hen the need for variety has historically been less important than it be-
10
comes,
standardization is the status que.
so that
buyers,
Witli
more than two groups of
move by one cannot necessarily be simply described as towards "more"
a
or "less" standardisation;
the matter then becomes even more
complicated".
Excess Inertia and Excess Momentum
3.
There can be
naive sense,
a
this is
substantial loss
com.-non:
art (we would not choose
froir,
being on the wrong standard.
many standards are
inferior
to
ther if we were designing afresh).
the
It
state
is a
In a
of
the
more dif-
fir_lt question whetiier the benefit to sv.-itchmg exceeds the transition costs.
Suppose
ter,
that
there
is
a
status-quo standard,
technology appears on the scene. How
v.'ell
and
a
new,
possibly bet-
does "the market"^*
solve
tiie
probiex of whether to switch, and if so how quickly?
In
\.'ill
net
tia". We
Ferrell and Salcner
s'..'itch
(19cSa,b)
even though it should.
we
showed that sometimes
We
called this effect
market
the
'excess
alsc discussed the opposite phenorr^encn of wrongly abandoning
iner-
a
tech-
v.'here
there
nology, which we called "excess momentum".
J.
7
Coordination Problems
In movies of the old West,
were nc trees
another.
Even
to
cowboys who camped for the night
which tc tie their horses would often tie the horses to one
thoual'i
the
hcrses
as
a
aroup
were
free
to
ao
wherever
they
11
go far -- whereas
they would not
wanted,
single
a
horse left free overnight
difficulty in coordinating on just where they would move at
would. The horses'
any instant prevented them from moving effectively.
In much the
an
same
and inferior
old
move to
happen
it can
v;ay,
even
technology,
v;lien
industry may get stuck on
tliat
an
all
participants might prefer
new technology. This happens because the group is "tied together" by
a
reluctance to sacrifice the benefits of being compatible. The fact that
not only horses who have
half
a
this problem is
dozen people walk from office
than witli
a
shown v;hene\'er
restaurant:
to
erm.ined order)
model
a
m
v.'hich
to sv.'itch to
a
each of
a
each user has preferences over its
ov.-n
externalities" assumption:
prefer others to make the same choice.
(1955a)
that,
all users
if
they will all s\:itch
erences differ,
outcome,
group of more
progress is
far
is
than
slower
number of users chooses (in predet-
new technology or to stay with the old. Because
of network externalities or other benefits
"netv;ork
a
it
smaller group.
Consider
a
to
of standardization,
that
choice and others' choices that satisfy
v;hatever
VJith
choice
a
user makes,
it
will
we showed
complete information^'
would be better off on the
(in the unique
we assume
new technology,
then
perfect Nash equilibrium). If their pref-
then the early movers have considerable pov;er to determine the
because
of
the bandwagon
effect.
This
result
called "the
(which we
New Hampshire Theorem.", from the timing of political primaries) comes from our
user has only one chance to svntch;
assaT.pticn that
a
are Stackelberg
leaders.
More
realistically,
v.-hatever
thus,
makes
the early movers
a
user able
to
commit itself early to a decision on standards v;ill give it pov/er.
However,
fectly
knov.'n,
v;hen
v;e
and studied
allowed for
a
the
model in V7hich
fact
that preferences
eacli
user could choose to sv;itch or
are
not per-
12
we showed that
each period,
not at
prefer the eld teclinology
the benefit
but none switch.
,
of the network
m
would be followed
a
there can be
symmetric excess
With incomplete information about
externality to others,
s\.-itch
inertia: all'
no user
can be sure
that it
This uncertainty can lead
to the new technology.
all the users to remain with the status quo even wlien they do all
m
fact fa-
vor s^.'itching, because tliey are unwilling to risk switching v.-ithout being follov;ed.
Non-bmdmg communication
that possibility,
user
Thus,
tlie
and inteiitior.s
there
be
car.
the
other would be
socially excessive reluctance
Similarly,
"excess
eliminates
asymmetric problem (if one
off if both switched, but
standard.
established
eld
actually exacerbates
it
be much better
v.'ould
what v;orse off).
an
but
about preferences
momentum"
is
to
some-
abandon
possible:
all
users may switch, even thougii it v;ould be more efficient not to dc so.
Thus
It
takes
a
substantial preference
for
the
standard to be
nev;
ari
early adopter; and if there are no early adopters then the standard will never
be adopted.
cut en
a
Excess inertia arises
lirJc
v.-hen
not
enougii users are
keen eiiough to go
with the new technoloqv.
3-2 Installed Base Problems
I.n
the r.cdel ^ust
standard is that
a
ing is more easily
ing back
etc.). We
IS
seer,
no-..-
described,
the
"switch" is more of
c.-.iy
a
as\'nn£try betv.-een
commitment than
reversible (by s\:itching later)
as an adr.ission
(on the cid technology)
is
"stay": not switch-
than is sv.-itchmg (switchv;ill be
very displeased,
that arise v;hen the
installed base''
of error,
sketch some considerations
a
the eld and new
customers
important. Tne analysis is
from,
our 19£5b paper.
13
With
installed
an
base
externalities are important, it
perior that it
will.
On
the
v.'ill
pay for
other hand,
that makes
adopt it,
v.'ill
the
eld
technology,
and
network
if
be rare for a new technology to be so su-
user to adopt it if he believes that nobody else
a
if
using
he
believes
everyone else from now on will
that
more attractive and also makes the old standard less
it
attractive. Technically,
that there may v;ell be two different ful-
this means
filled-expectations equilibria. What happens depends purely on what people exto happen.
pect
instance
ures
Naturally,
by strategic
this
encourages manipulation of expectations,
forecasting,
boasting,
exaggeration
or
of
sales
for
fig-
.
Although the case of multiple equilibria is common and important,
analytical conclusions are more clear cut if there is
a
the
unique equilibriums We
can show that there can be a unique equilibrium in which the new technology is
ignored but should be adopted,
and that
equilibrium in which it is adopted
The
reason for this
ternalities.
adopters.
that
each
First,
Second,
adopter
affects the value of
adept the new technclog-y.
Katz and Shapiro
there can be
a
unique
should not be.
the values of
individually may be
an installed base but
it
other cases
that each user's
affects
it
it
is
v.-hen
in
the
the
either
adoption decision has two exoptions available
installed base.
too
keen
or
too
The
to
future
result is
reluctant
(19c5b)
study the questions
that
arise when there is
also "sponsorship" of one or both technologies,
so that
sellers may engage in cross-subsidization between early and late users.
shov;
dard,
to
^
that the market outcome may involve standardization on the
They
"wrong" stan-
and may standardize when there should not be standardization.
14
4.
Anticompetitive Practices
Are
anything,
if
What,
anticompetitive
theie
should be done about them?
preliminary analysis on
The power
</./
smaller
the
it;
does net
fir.-r
tliese
tlie
care much
firms
standardization game.
whether the
more
the
industry.
If a
miay
Barrier
m.cdel,
able
tc
by
able
to ignore
geous terms
m
veloped
-
m
a
firm is
com.patible
the large firm, can unilat-
Both lEM and AT&T are widely believed
advantage is
render it valueless by
not
itself anticompetitive,
but
it
We discuss three belov;.
others'
buying
refusing to adopt it.
would be cutting its
firm,
such behavior;
run)
that
Entry
large
the
position
larger
firm or an industry outsider develops an imprcve.ment
sm.all
be
to
reason is
The
smaller rivals are compatible with
its
does open up possible anticom.petitive strategies.
^2
section we report some
this
tne possibility arises that
This
ability.
this
In
standardization'
to
questions.
care mucli wiiether
erally set standards for
have
m
power
with them. Accordingly,
to
related
of siie
Size confers
large
practices
but,
with important
im.provements
them..
,
own throat
net\.'crk
In
large firm
a
stand.= rd
the
m
least
externalities,
or alternatively
the
it
long
may be
negotiate very advanta-
For example. Western Union had a strong bargaining
buying telecomm.unications patents before
though not
^at
,
so strong
as
it
thought
v.-hen
it
tne
Bell
system,
\-ias
de-
turned down an offer of
15
Bell's telephone patent for $100,000. Obviously, this effect tends to maintain
market power,
potential and actual competition.
and reduces
the network externality creates
a
In
other words,
barrier to entry. This may well be the most
important anticompetitive effect from standards.
discuss some other
Below, we
possible effects.
^.3 Predatory Standard Switching
If
a
firm
that
will suffer from making
rivals. Wherever
a
can set
de
sv.'itch
a
in standard,
firm can hurt its
(if re-entry is difficult)
dation
facto standards
rivals,
perceives
that
its
rivals
then its way is open to harm its
there is the possibility of pre-
and of raising rivals'
costs (see Salop and
Scheffman, 19S3) in any case.
Thus
is a
suppose
that
firm A unilaterally changes its
standard.
Because A
dominant firm, buyers expect its new standard to be the standard for the
future.
However,
to exit
Therefore,
firm
this may cause
the m.arket,
has
no
option but to
switch to A's new
disruption of B's activities.
then A has effected
Even if B does not exit,
ened^'.
B
a
If B
is
standard.
thereby caused
predatory reduction in competition.
its competitive position relative zo A may have
v.-eak-
16
Product Announcements
ii-U
m
One of the alleged anticompetitive practices named
partment
against
suit
IBM
v;as
"premature"
hov/
ticompetitive. One would expect (see Fislier et al.,
of
a
such
a
However,
market's
practice can be an-
19S3) that an announcement
ve
showed
externalities
networl;
to
sclutic.-.
19c5b paper
our
iri
(see
also Katz
these views
important.
are
question of whether
the
necessarily cptir.al
that both
and
to
The
adopt
Shapii-o,
pred.^.tcry
m
the
sense
ficing short-run profits
when It succeeds
m
reasonable to expect
blocked,
net present.
sc
It
that
is
ir.
dcmg
v.-hen
the
alsc
that
the firm,
that
19£5b)
uiidertakes
are potentially
reason
the
inforr.otior. does not necessarily lead to better outcomes.
be
a
announcement of an inferior product would have no effect.
an
v.'hile
misleading if
be
In
'
superior product would be socially beneficial (though detrimental to com-
petitors)
not
Justice De-
announcement ^ '.'
product
standard economic frame-oork, it is hard to see
tlie
new
it enhances
its
that
tlie
technology is
and so
the
extra
The announcement can
the action
order to cause the exit or failure of
so
is
a
is
sacri-
rival,
future profits, tioreover,
it
and
is
installed base is important that future re-entry may
usual problem
•-•crth
m
explaining what can be predatory is
noting that the standard tests for predation may
fail to detect the predation. See cur 15£5b paper.
17
5.
Incentives for
R&D
5.1 Excess Inertia
and R&D Incentives
How does the possibility (discussed above) of excess inertia affect the
incentives
glance,
it would seem
R&D will be
that
v;ards
for R&D
means
that part of
are
v;ill
less
the
development
(R&D)
of
new
Even if
be adopted
firm develops
a
m
the market;
than they would be without
a
tia
At
better system,
this
first
it
that the
im.plies
excess
inertia.
This
is
re-
also
inefficiency caused by excess inertia is invisible to
even the most careful study of inventions not adopted,
R&D projects never get
technology?
influence of excess inertia on incentives for
the
straightforward.
certain that it
not
and
research
for
started (or pursued)
as
a
since presumably many
result of anticipated iner-
.
However, that argument is based on considering only one potential inno-
vation.
In
return
the
fact,
to
any innovation
is
limited by
something better still will come along; and this leads to
fication.
In
simple model^^ we can show that having
a
minimum successful improvement on existing
pected returns
helps
him.
to R&D.
Intuitively,
The help is
later.
the
time
v.'hen
substantial modi-
higher
threshold of
technology indeed reduces the ex-
barrier
therefore
a
a
the
harms an innovator
discounted,
and v;orth less
now,
and
than the
harm now.
R&D
v.'ill
fall over time,
be
especially discouraged
v;hen
excess inertia is expected to
so that the barrier protecting current incumbent technology is
higher than that expected to protect anyone who overcomes it.
As
the
the
"base"
incentive for R&D
grows over
v.'ill
be
time,
inertia will tend to
increase^'.
Thus
strongest early in the industry history before
18
the standard gets locked
taking over may be
former
m.
Also,
the best time
the time
when
to introduce
new technology seems
a
further innovation,
a
to be
since the
nc installed base.
lias
5-2 Network E xternolities and Licensing of Patents
In
a
market with important network externalities,
centive to licence
pater.ted or sponsored
a
technology.
to keep liis new technology to i'.imself.
in\'entor
there is an extra innot pay
does
It
it may pay to licence
Indeed,
at no charge to some buyers withi large volumes but low v.-illmnness
For example,
cor-.puter
for an
to pay.
makers have been very willing to donate machines
to schools and universities.
Another example: Xerox licenced its local area netv;ork "Etiiernet" at no
cost to other suppliers.
become
edge,
make
standard,
a
and
the industry
or
tiiat
is
tiie
originator will
l-.ave
sufficiently unccT.petitive
satisfactory prcfit
a
presumably intended to ensure that Etiiernet would
It
fro.T.
its
sh-are
of
,
enough of
the originator
that
the market
cor.petitix'e
a
for
v.-ill
standardized
the
product
An clternati\'e strategy of keeping the technology proprietary was followed
earlier
though
c
a
s-all
dard.
On
ir.
same
r.arket
by
Datapcir.t,
proprietary standard right have
firm such
the
less reason
as
to
been
beer, a
Datapomt had difficulty
ether hand,
Xerox would have
market
the
of
course,
difficult for
believe it could
m.ake
the
•.-.•hich
was
unsuccessful.
viable alter.-.ative
m
establishing
it
nonproprietary strateg-y
Datapomt
too,
since
satisfactory profits in
for /lerox,
as
a
stan-
followed by
right have
it
a
Al-
had
nonproprietary
19
5-3 First
Mover Advantage and
nnovation
I
The first firm to introduce an innovation often retains
(price premium, or ability to maintain
tage
others have entered and imitated.
competition,
as
it gives
V^iile
a
a
market advan-
large market share)
tins is
in
even after
some sense bad for static
monopoly pov;er to the first mover,
(limited)
been argued that such first mover ad\-antages
may enhance dynamic
peterian" competition by providing something akin to
or
it
has
"Schum-
patent reward for inno-
a
vation.
We can consider how
inertia
cess
m
a
first mover advantage affects the analysis of ex-
19S5a paper.
our
advantage encourages innovation.
together v;ith
less
likely to be followed in
IS
desirable
if
probability that
need net be
others follow,
others
will
to new
sv;itch
a
but not
follow
is
if
case
the
The reason is that
second mover disadvantage,
goes
a
It
that makes
technology.
critical
do not.
to
his
tive,
f ollov.'ed.
This makes
the
decision.
vantage need not encourage innovation.
case without network externalities,
part of rivals merely improves
tor.
)
m
perceived
A
second-
This in turn
innovation less attrac-
and this indirect effect can outweigh the direct effect of
advantage, which is to make innovation more attractive.
innovating
Often,
Thus
advantage
first mover
a
mover disadvantage makes following an innovator less attractive.
reduces the probability of being
first mover
a
the first-mover
and
they
that
Thus,
a
first-mover
firsi-mover ad-
(Note the contrast with the "standard"
which any reluctance to im.itate on the
the competitive
position of the first innova-
20
6.
Conclusion
Standaidization
m
the
extremely important
is
information processing industries.
m
modern economies,
VJhile
it
lias
especially
many benefits,
it
may
also have serious social costs. There has been little economic analysis of the
policy problems
Conclusions reached by traditional economic reasoning,
ity and diminislimg returns are generally assumed,
when bandv/agon effects,
problems are important.
v.-indov.'s
IS '/ery different
There
is
which
netv.'crk
other
v.'ork,
mind,
a:-.d
we
no easy
are likely to be misleading
opportunity for entry,
v.'h.er.
predation, R&D,
and the
have
have
licencing cf inven-
prescription for m.icroeconomic policy
identified
shov;n
and installed base
compatibility is important.
externalities play an important part.
we
wluch convex-
We have seen that the analysis of sucli staples of in-
dustrial orgar.izaticr. as pricing,
tions,
of
m
hov:
treated ceutiousiv. Furt'ner
some
certain
\:cvV.
of
the
factors
standard
In
this paper
that
lessons
in markets
of
on the subiect is needed.
and
in
m
our
should be
kept
m
economics
m.ust
be
21
Footnotes
For
(1985).
1.
a
general
discussion
of
standards
in
telecommunications,
see
Berg
is sometimes possible to find translation protocols, which, if sufficiently effective, render moot the question of standardization. Such translation, however, is sometimes impossible and is often so difficult that standardization remains important.
2. It
Cost reductions may be valued by users or by makers (or both), depending on
economists call the incidence of cost savings, which is a matter of market structure. For example, in a perfectly competitive market, all cost savings that reduce marginal costs are passed on to buyers, but cost savings that
reduce fixed costs but do not affect marginal costs are not. In imperfectly
competitive markets, some cost savings may go to enhance the profits of sellers
3.
v;hat
.
For example, Fisher (1979) v.'rites
To the extent that certain boots are associated with certain bindings- -the real competition tal<e5 place between binding-boot combinations-
4.
5. We mean this term in a broad sense, to include not only computers and their
peripheral devices, but also cameras and lenses, ski boots and bindings, nuts
and bolts, autos and fenders, etcetera.
6.
See for instance Farrell,
1985.
This raises the question of why sellers might resist standardization, given
that it enhances the value of their product and also helps tov/ards an efficient selling system. The ansv;er may lie, as Katz and Shapiro (1985a) suggest,
in some sort of costs of standardization. Perhaps more plausibly,
as Berg
(1985) has suggested, standardization somehow makes competition more direct.
Thus firms trade off the incentive to maximize the value of their product
against the desire to restrain competition. We are also working on this line
of inquiry.
7.
If £ seller is large enough, there may be a presumption that someone would
take over these support services, but it is typically not the large sellers
that suffer from this fear.
8.
In markets with a high (or potentially high) degree of vertical integration, moreover, buyers may not v.'ish to be strategically dependent on a supplier v7ho may later become a direct competitor. Sometimes a user may avoid a
de facto standard for this reason-, the user fears that the de facto standard
setter may exploit his position. Some of the standards decisions of the European community reflect such a concern. For an interesting history of European
standards decisions on color television, see Crane (1979).
9.
Indeed, some antitrust decisions have suggested that an ovjner of a netmay have an obligation under law (as opposed to regulation) to allow interconnection by competitors, if the network becomes an "essential facility".
10.
v7ork
22
which it sometimes is, a seller v,'ould
11. If It IS technically feasible,
always wish to have one-way compatibility: his buyers get the advantages of
the whole network, while other firms' customers only get the externalities associated with their small networks. A classic case of this is the competition
tlie safety razor blades that dominated
that led to the strange-shaped holes
the market until the seventies.
m
12.
For example,
see Dixit and Stiglitz
(1977),
or Spence
(1976).
13. For a discussion of the possibility of compromising on the
Berg (1955).
standard,
see
m
tastes over
K. For another analysis of standardization witli differences
standards, see Berg (19S5). The work of Katz and Shapiro (1955b), and Farrell
and Saloner (19S£b), althougli not explicitly about the efficiency-variety
tradeoff, can be interpreted in those terms. They analyse the evolution of
standards over time v.'hen buyers at different times have different preferences
or different opportunities.
We focus on what Artl'iur (19S4) has called "non-sponsored" technology. That
there is no ov.-ner of the n.'w (or the old) standard who car. subsidize early
adopters
order to get a new standard rolling.
Interesting questions arise
(Katz and Shapiro, 19£5b, Hanson, 19=5) if the technology is sponsored, and
also if, while the new standard v.-culd not be owned, its early developer v.-ould
have a substantial first-mover advantage if it succeeds. (We briefly treat tne
question of first-mover advantage below.)
15.
IS,
m
Ic.
ers,
This term means that eac'r. decisionmaker knows the preferences of all othand everyor.e knov;s that, etc.
m
17. Sometimes the installed base \.-ill be physical capital,other cases the
human capital invested
training on the old system will be more im.portant.
The tinpevTiter keyboard case (David, 19£5a) is an interesting intermingling of
the two
m
IS.
n related problem (without the dyna-.ics) is the problem, cf uneconom.ic bypass
telephone m.arkets. It is v.-idely predicted that, if pricing is constrained by uniformity requirements, bypass will become very popular and will
reduce the value cf being on the netv.-ork (here, for cost-sharing reasons) to
such an extent that the bj^pess is socially inefficient. Again, an autarchic
technology- m.ay reduce the value obtained from a netv.xrk technology; agair.
this v.-ould not happen if pricing voere not constrained.
m
not speculate here on v;hether any particular standard switch is
However, Brock (1975) describes the announcement of IBM's System
360
terms that suggest such an interpretation. Honeywell had announced a
computer known as the H-200, together with a translator device (called the
Liberator) making it compatible with the IBM 1401. This combination competed
directly and favorably v.-ith IBK's 1401 machine, v;hich was then the market
leader.
IBM's System 360, v/hich was announced shortly after Honeywell's announcement, was incompatible v,'ith the 1401 and the H-200, and yet (whether because the 360 was superior, or because of IBIl's reputation as a de facto stan19.
We
v.'ill
predatory.
m
SJi^dard setter) customers chose to
commercial viability of the H-200.
wait
for
the
'
—
360,
-..
--
thus
•"'.
destroying
23
the
See allegations in California Computer Products Corp. v IBM Corp., 613F.
727 (9th Cir. 1979), Memorex Corp. v. IBM Corp., 636 F. 2d 1188 (9th Cir.
1980), and see also Berkey Photo Co. v. Eastman Kodak Co., 457F. Supp. 404
1093 (1980).
(5.D.N.Y. 1978), 603 F. 2d 262 (2d Cir. 1979) and 444 U.S.
20.
2d.
IBM's rivals have also complained about insufficient advance notice of
Evidently, firms that
IBM's changes in its (and de facto industry) standards.
depend on following the lead of a de facto standard setter are at a competitive disadvantage if they learn about standard shifts only late in the day.
IBM has agreed to announce standard shifts in advance, as a result of an EEC
antitrust action.
21.
22.
Details available on request.
23. This may account for some "waves of innovation": no sooner does innovation
A start to succeed than innovation B pushes it out of the way; in contrast to
other industries in which there has been no innovation in many years. It may
also account for the tremendous economic growth sometimes observed after wartime destruction of the capital stock.
Once a large fraction of the existing
stock is destroyed, there is (or can be) a general expectation that new techniques can profitably be put in place. See Shleifer (1985) for a business-cycle model related to this observation.
.24
References
Arthur, E., "On Competing Technologies and Historical Small Events:
The Dynamics of Clioice under Increasing Returns," mimeo, Stanford, 1983
Berg, S., "A Duopoly Model of Technological Externalities and Compatibility Standards," mimeo, University of Florida, 19S5
Brock, C, "Competition, Standards and Self-Regulation in the Computer
Industry", in R. Caves and M. Roberts, eds, Regulating the Product: Quality
and Variety. Cambridge, HA: Ballinger, 1975.
Char.dier,
A.,
The
Visible Hand.
Cam.bridge
:
Harvard University Press,
1977.
Crane, R., The Politics of International Standards:
Color TV War. IJcrvcod, K.J.: Ablex Publishing, 1979.
France
and
the
Dascupta. P., and J. Stiglitz, "Exercises in the Economics of Learning
by Dcmg," paper presented at the GTE Laboratories Economics Symposium, August
19E5
.
David, P., "Ciic and
view, 75, 332-23c, May 1955a
the Economics
of QV;eF.TY
,
American Economic
"
Re-
David, ?., "Nev Technclogy Diffusion, Public Pclicy and Industrial Competitiveness," Center for Economic Policy Research, Stanford, 46, 19E5b
Dixit, A., and J. Stiglitz, "Monopolistic Competition and Optimum Product Diversity," American Economic Reviev; 67, 1977.
Farrell,
vation'
J.,
"Cc.mpetition and Lock-In," m.imec,
GTE Labs,
Farrell, J., and G. Salcner, "Standardization,
Rand Journal cf Eccncr.ics, Spring 19S5^a)
1SE5.
Corpatibility and Inno-
,
Farrell, J., and G. Salcner, "Installed Base and Compatibility, V.'ith
Im.plications for Product Preannouncement " mimeo, GTE Labs and MIT, 19S5(b).
,
ing.
Farrell, J., and G. Salcner,
Economics Letters, IrSS^c).
19£5(d)
Farrell, J., and
in progress
19E5{e)
Farrell, J., and
progress
G.
Saloner,
"Standardization and Variety,"
forthcom-
Variety,"
mimeo,
Timing of Standardization,"
m.imeo,
"Monopoly
and
Product
,
,
m
G.
Saloner,
Federal Trade Commission
Staff Report," April 1953.
"The
(FTC),
"Standards
and
Certification:
Final
25
Fisher, F., "Diagnosing Monopoly," Quarterly
Business, 19, 1979.
Review
of
Economics
and
Fisher, F., J. HcGowan, and J. Greenwood, Folded, Spindled and Mutilated: Economic Analysis and the IBM Case, Cambridge: MIT Press, 19S3.
Hanson, W., "Bandwagons and Orphans: Dynamic Pricing of Competing Technological Systems Subject to Decreasing Costs," mimeo. University of Chicago,
19S5.
Katz, M., and C. Shapiro, "Network Externalities, Competitition
Compatibility," American Economic Review, 75, 424-440, May 19S5a.
v;ork
and
Katz, M., and C. Shapiro, "Technology Adoption in the Presence of NetExternalities," mimeo, Princeton University, July 1985b
Hemenway, D., Industrywide Voluntary Product Standards, Ballinger 1975.
Landes, D., The Unbound Prometheus:
Present. Cambridge University Press, 1969.
Technological
Change
1750
to
the
Landes, D., Revolution in Time. Cambridge: Harvard University Press.
Ordover, J., and G. Saloner,
T'lT, in progress.
"Predation, Monopolization and Antitrust,"
Salop, S., and D. Scheffman,
Review 73, 1983.
"Raising Rivals' Costs," American Economic
mimeo,
Shleifer, A
.,
Spence A.M.,
Review 66, 1976.
,
"Implementation Cycles," mimeo HIT 1985
"Product Differentiation and Welfare," American Economic
m^d^ss)
6340 u64
3
T06D DD56ETt.a
1
:-:m:-m?^ms^
^^^
\8tDue
MAY 01 1992
Lib-26-67
MIT
3
LIBRARIES DIJPL
TDflQ
1
DD5fl2Tt,a
1