Mil. LIBRARIES - DEWEY
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium
Member
Libraries
http://www.archive.org/details/onmanagementofinOOaghi
0£\M£Y
$'
-«B31
.M415
l~2-
working paper
department
of economics
ON THE MANAGEMENT OF INNOVATION
massachusetts
institute of
technology
50 memorial drive
Cambridge, mass. 02139
ON THE MANAGEMENT OF INNOVATION
Philippe Aghion
Jean Tirole
No.
93-12
July 1993
TufTUBRARiES
DEC
-
7 1993
RECEIVED
On
the
Management
of Innovation
Philippe Aghion
Jean Tirole *
5
t
July 23, 1993
*We thank Patrick Bolton, Fernando Branco, Drew Fudenberg, Oliver Hart,
Rebecca Henderson, Gary Pisano and Diego Rodriguez for their help. The second
author is grateful to the Commissariat General du Plan for financial support.
^Nuffield College, Oxford and European Bank for Reconstruction and Development, London.
*Institut d'Economie Industrielle, Toulouse, MIT and CERAS, Paris
Abstract
Empirical and theoretical analyses of the management of innovation must carefully
entangle its several dimensions:
over the
Who
R&D process and property
finances
R&D
expenditures?
rights over the innovation?
Who
dis-
has decision rights
Who receives the return
streams
derived from the innovation?
This paper
for
is
a
first
attempt at opening the
R&D
blackbox.
co-financing by investors and users of the innovation and
shop
rights, trailer clauses
and the hired-for doctrine.
of Schumpeterian hypotheses.
joint ventures.
Last
it
It
for
It
studies the foundations
employment
features such as
investigates the theoretical validity
analyses the incentive costs and benefits of research
Introduction.
1
Following Schumpeter's (1942) persuasive case that
R&D
is
a major determi-
nant of economic growth and welfare, empirical research in industrial organization has investigated the sources of innovation and their incentives.
It
has
extensively tested hypotheses enunciated by Schumpeter and his disciples,
that a firm's incentives for
monopoly power and
initial
large
R&D
body
would be higher, the larger the
financial resources of the firm.
R&D
The
resulting
ambiguous conclusions concerning these
of evidence has yielded
determinants of
scale, scope,
(Cohen-Levin (1989)). Another strand of the
ture investigates the nature of research rather than the
R&D
litera-
effort or its
output. For example, empirical work has started to test (with mixed results)
the idea that established firms would pursue incremental rather than radical
innovations (Henderson (1991)).
Surprisingly, little attention has been devoted to the theoretical validation of these hypotheses.
the organization of the
at several levels:
Nor has one more generally derived a theory
R&D
activity
1
.
Theoretical foundations are needed
to interpret existing findings, to suggest
testing, to build efficient private organization of research,
ernment intervention
offer
a
first
in the
must
market.
Who
its
There
is
and to guide govof this paper
is
to
finances
R&D
rights over) the innovation?
Who
three dimensions:
owns (has property
receives the return streams derived
1
designs for
and theoretical analyses of the management of
carefully disentangle
expenditures?
to license.
The purpose
new
step in this direction.
In our view, empirical
R&D
R&D
of
Who
from the innovation? To analyze these
a large theoretical literature on the study of patent races and incentives
focus of this literature is the persistence of R&D and product market
The
monopoly power and not the management of innovation.
distinct dimensions, let us note that research
is
usually not performed by a
firm, but
by a person or group of persons, the research
that will
become
clear later, the
non research parties
unit.
For reasons
also involved in the
organization of research must be divided into two categories: customers, and
financiers or investors.
fit
Customers are those parties that
from the innovation; namely, the manufacturers who
the innovation, the users
will
will
commercialize
purchase the resulting product, and the
complementary products or of inputs used by the manufacturer.
suppliers of
(Which
who
will directly bene-
of these three kinds of customers finances the innovation
depends on
the industry (von Hippel (1988)) and will only be briefly discussed in the
paper; for
now we can content
The customer and
firm.
It
is
ourselves with a single aggregated customer.)
the research unit
may
interesting to note that, in
customers spend a large fraction of their
R&D
may
or
some
R&D
not belong to the same
sectors such as biotechnology,
budget on sponsoring external
2
.
In contrast to customers, investors benefit from the innovation only to
the extent that they share the royalties from the sale or licensing of the
research unit's patent or trade secret. Again there are several categories of
banks or venture capitalists who acquire shares or other securities
investors:
in the research unit;
that
own
and parent companies (including research companies)
the research unit and operate in an industry or product line different
from that of the customers, and therefore
innovation
will
not benefit directly from the
3
.
The relative proportion of in-house R&D and externally sponsored R&D is not always
known with precision, as data often record only the total R&D budget. Pisano, Shan and
Teece (1988) for example report that Monsanto spent in 1982 $40 million out of an R&D
2
budget of $62 million on outside contracts.
3
One may rightly argue that the distinction between customer and investor is less clear
cut than it appears. An investor might enlarge the scope of its activities and start commercializing the innovation. The theory developed below is straightforwardly reinterpreted to
There
is
a wide range in the distribution of property
instance the case of integration in which the research
the customer or another firm.
The employee
The
research unit
is
rights.
is
Consider for
performed within
then an "employee" 4
relationship does not quite imply that the employer receives
property rights on
all
employee-generated inventions. Employment contracts
as well as the law relate ownership to the nature of the invention.
variations, property rights normally go to the employer.
was "not hired
tion that the employee
for"
for
the employer's
With some
However an innova-
and which was performed outside
the firm usually belongs to the employee. Innovations that
namely those
.
fall in
between,
which the employee was not hired, but that made use of
facilities or
employees, exhibit
split
data or benefited from the assistance from fellow
ownership
The employee owns the
rights:
innovation,
but the employer has a "shop right" to receive a nonexclusive, nonassignable
and royalty-free
license
5
.
A
related, but different allocation of property rights
confers ownership to the employer for inventions that are "in-line",
relate to the employer's business,
and to the employee
of innovations
made during
and those made long
among
inventions
is
that
for inventions that are
"not-in-line" in the sense that they pertain to another business.
poral version of such distinctions
i.e.
An intertem-
the different treatment
or just after the employee's tenure in the firm
after the breakaway.
Another important dimension of R&;
D
contracting
is
the distribution of
An "investor" who may or may not benefit directly from the ina combination of an investor and a customer. Von Hippel (1988, p45) suspects
that there is little role switching in practice.
allow for this possibility:
novation
is
4
Incidentally, Schumpeter (1942) and Galbraith (1952) were struck by the fact that
innovation often takes place in R&D labs administered by large corporations. Indeed,
Stedman (chapter
essentially
2 in
a personal
Neumeyer (1971)) notes
act, 60
%
that, despite the fact that creation
is
of the patents issued by the U.S. Patent Office in 1954
went to corporations, as opposed to 35 % in the mid 30's.
5
There are some exemptions to this pattern. Neumeyer (1971) records that, unlike
the three other big US aircraft manufacturers, Boeing permits its employees to retain
ownership of their inventions provided Boeing receives a shop right.
return streams between the research unit and the customer or investor,
the innovation
is
when
sold to third parties (other customers). Venture capitalists
routinely take shares in high tech research units.
And many
corporations
specify the division of return streams for innovations generated by their
ployees and sold to third parties
6
.
The sharing
7
covary perfectly with the property rights
is
.
em-
of return streams does not
This does not
mean
that there
no relationship between ownership and royalty sharing, as both should be
determined simultaneously as part of an optimal incentive package8
.
This paper develops a simple model that sheds light on static and dynamic
configurations of property rights and sharing rules in research organizations.
The paper
is
structured as follows: Section 2 analyzes the basic relationship
between a research unit and a customer (manufacturer, user, or supplier of
related components).
The customer
the only potential customer or else
is
signs an exclusivity contract with the research unit.
The
research unit per-
forms the creative task but has no independent resource to pay for
equipment or data.
step,
we assume
It
must therefore look
that the financing
is
6
We
is ill
posit
defined ex ante and that the
US aircraft manufacturers offer their employees shares
on income collected from royalties (Neumeyer (1971, chap. 4)). Some
For instance, the four largest
of 10
%
to 30
%
universities give 15
7
In a first
for outside financing.
provided by the customer.
that the exact nature of the innovation
salaries,
% of the
royalties to their researchers.
For example, property rights go to the employee in universities and to the employer in
manufacturers even though the sharing rules are similar (see previous footnote).
in point is that venture capitalists or customers who purchase a majority
of shares in a venture may not ask for control rights (The Economist, August 29, 1992,
aircraft
Another case
p53-54).
When
Hoffman-Laroche bought a 60
%
stake in Genentech in 1990,
it
demanded
only 2 of Genentech 's 13 board seats.
8
Furthermore, the very feasibility of royalty sharing may depend on the ownership
In the above aircraft manufacturers example, the manufacturers sometimes
structure.
expropriate the employees' share of royalties by granting each other free licensing of their
This expropriation would not occur under employee ownership. In another
by its owner,
the US government forces its employee-owners to exploit their inventions diligently through
compulsory licensing.
inventions.
illustration of the difficulty in guaranteeing the intended use of the invention
two parties cannot contract
tract specifies a verifiable
for delivery of
amount
The
a specific innovation.
con-
of customer investment, the allocation of
property rights on any forthcoming innovation and, possibly, a sharing rule
on the
profit (license fee)
obtained by the research unit 9
In the integrated case, the customer
owns and
In the nonintegrated case, the research unit
the innovation
is
.
freely uses the innovation.
owns the innovation and, once
made, bargains with the customer over the
sharing rule contracted upon ex ante
is
shown
to
be
license fee.
irrelevant.
The
The study
then boils down to the classic one of choosing property rights so as to best
10
protect the two parties' specific investments in the relationship
property rights to the research unit
encourage the unit's
cial
is
optimal when
effort to discover
it is
.
Giving
more important
to
than to boost the customer's finan-
(and nonfinancial) investment in the research, or when the customer
is
subject to financial constraints (the long-purse effect).
Our framework
hypotheses.
It
also provides a rigorous evaluation of the
Schumpeterian
questions the robustness of the Schumpeterian conjectures
mentioned
earlier to variations in the allocation of property rights
novation.
Once
this allocation
is
on
in-
endogenized as part of the contractual
arrangements between customers and research units, one should not expect
a clear empirical aggregate relationship between
parameters such as
Is it
9
We
scale,
R&D
input or output and
scope or monopoly power.
optimal to have a single providor of funds, namely the customer? In
can allow the research unit to own some shares of the customer stock or
As discussed
in section 2, this
profit.
hardly affects the analysis.
10
It will be obvious to the reader that our basic model borrows considerably from the
property rights analysis of Grossman-Hart (1986). There are a couple of differences in
the treatments. First, we allow equity participation by the trading partner as well as by
third parties. Second,
we introduce
contractible (as well as noncontractible) investment
and point at the role of cash constraints. These features account for a couple of results
of independent interest such as the rationale for cofinancing or the possible inefficiency of
the allocation of property rights.
we show
section 3,
that
it
may be
strictly
property rights to the research unit and to
optimal for the customer to give
demand
cofinancing by an investor
(such as a venture capitalist, a research firm or a parent
customer's business).
The
The
investor then takes
research unit will resist cofinancing as
it
some share
of cofinancing
n
if
in the
in the research unit.
dilutes its share of the pie, but
can only consent to give shares to an investor
position ex ante or else
company not
if it is
in a
weak bargaining
the customer does not break even in the absence
Furthermore, this arrangement cannot be duplicated by
.
transferring the investor's share in the research unit to the customer, since the
customer then faces conflicting objectives when bargaining with the research
unit.
We
thus obtain a theory of the existence of multiple. principals.
Section 4 considers product market competition
particular analyzes the validity of the
search units have
division of
more
common
among
customers.
claim that independent
an incumbent customer.
We
study a choice
and making a more drastic discovery.
unit's preferred choice of technology
pay
among
research fines
We
that independent research units have
more incentive
there
to
is
no presumption
make
drastic innova-
Bertrand competition in the product market,
it
less incentives.
Shop
and
11
show that the research
depends on the relative willingness to
We conclude that without more specific assumptions,
tions. Indeed, in the case of
(or
incumbent customer and of potential entrants.
for the innovation of the
has
re-
incentives to pursue radical research than a research
and analyze the tradeoff between innovating with a higher probability
faster)
in
It
rights,
property rights contingent on the nature of the innovation
rules governing
On
breakaway research are instances of multiple,
the other hand, investors
may
split
prop-
not want to fully finance the research, as part of
the value of innovation goes to the customer.
erty rights. Section 5 rationalizes these institutions
by introducing multiple
customers or multiple innovations. The allocation of property rights between
the customer and the research unit
centives to both parties.
The
is
then designed to provide maximal
basic principle
party on those activities for which
it
is
in-
to give property rights to a
has a comparative advantage in creating
value.
The
discussion of breakaway research in section 5 leads naturally to a
study of the dynamic management of innovations.
Vertical research joint
ventures (RJVs), which constitute a substantial fraction of RJVs, are de-
signed to bring together complementary assets, usually research capacity
and manufacturing or marketing
12
.
Starting a vertical
RJV
consists for
the two partners in identifying a research project, creating a separate entity,
defining control rights over this entity, specifying (contractible) inputs,
and
distributing equity participations. In a static context with a single innovation
in sight,
an
RJV
vation, that
is
is
but a special (nonintegrated) way of managing the inno-
of allocating the financing, the control rights
and the
profits.
Section 6 follows the business economics literature by stressing the specific
dynamics of the RJV. Indeed an
in
scope and/or in time.
It
RJV
has a specific objective and
By
The short-term
matters when the research unit cannot protect
releasing
its
technological knowledge to
its
We
horizon of the
its intellectual
property.
partner, the research unit
raises the probability of success of the joint venture,
12
limited
thus pursues some current innovation and keeps
the partners independent for future activities.
RJV
is
but also creates
its
own
RJVs: unrelated RJVs
two customers serving different product markets join forces to develop a common
input) and horizontal RJVs (the customers compete on the same product market). Horleave for future research the analysis of two other kinds of
(at least
izontal
RJVs
are of course subject to the strictest antitrust scrutiny (antitrust tradeoffs
and the Journal of Economic Perspectives symposium on the topic (1990)). Also some of the large horizontal RJVs exhibit a substantial
government involvement.
are discussed in Katz-Ordover (1990)
competition
in
the future.
RJVs
are thus not very conducive to technology
be softened by letting the cus-
transfers, although future competition can
tomer take an equity participation
in the research unit. Vertical integration
reduces the research unit's incentive to hold back the transfer of knowledge
by reducing
tomorrow.
payoff from having exclusive knowledge of the technology
its
We
show however that integration imperils future technological
progress.
Last, section 7 summarizes the
main
insights of the paper,
some
directions for research.
2
The research unit-customer
research unit
(RU) performs
innovation for the customer
is
by C. The probability
also
assume that
productivities of effort
is
l in
be
0.
So
is
the
and investment at
value of
probability of discovery, p(e,E),
by
RU
and on the investment
their zero level
minimum
(e,
We will
E).
its
level of
is
infinite so as to
level of effort of the research unit,
researchers' intellectual curiosity,
and prospects of informal rewards,
minimum
The
the relevant range and that the marginal
the level of effort induced by
ego, career concerns
The
effort cost e
guarantee interior solutions. The
that
0.
increasing and strictly concave in
is
E) <
p(e,
research for a customer (C).
V >
depends on the noncontractible
E
relationship.
Research Technology.
2.1
A
and discusses
is
normalized to
investment by the customer.
We
make two opposite assumptions concerning the customer's investment.
the
first
case
E
is
monetary and
contractible.
In the second case
E
will
In
stands
for proprietary technological information freely supplied to the research unit,
or for interaction with the research unit to tailor the innovation to the final
E
demand;
will
We
then be assumed to be noncontractible.
would of course
expect a mixture of contractible and noncontractible investments
in reality.
two cases are most often identical and so we
will state
The
results for the
the results with both cases in mind, unless they
differ, in
which case we
will
note the points of departure.
Without
p(e,
E)
=
loss of insights,
q(e)
we
posit a separable
form
for the technology:
+r(E). Our theory can be straightforwardly extended
separable technologies; the
new
of the customer's investment
feature
(if it is
is
to non-
then that the optimal specification
contractible) reflects its influence on the
research unit's effort through complementarities or substitutabilities in the
production function.
Both
parties are
income
risk neutral
and have reservation
utility 0. Fur-
thermore RU's income cannot be negative. Let us define the socially optimal
effort
and investment e*(V) and E'(V) by
m&x{p(e,E)V-e-E},
{e,E}
or
q\e*{Y))V
2.2
RU
= r\E*{V))V =
1.
Contracts.
has no
The two
money and
therefore investment
must be
parties cannot specify the innovation ex ante.
They can
allocate the
property right on the innovation to one or the other. Furthermore,
a (nonvoting) equity share
the licensing fee
13
13
.
We
(1
later
— a)
in
RU,
so that
RU
by C.
fully financed
C
takes
retains a fraction
show that one can choose a
=
1
without
a
of
loss of
In the whole paper we restrict attention to linear sharing rules for simplicity. However,
we shall argue in subsection 2.4, introducing non-linear sharing-rules will not modify
our main conclusions for the research unit-customer relationship, as long as the two parties
can by mutual consent tear up the initial contract and renegotiate.
as
generality. [For expositional simplicity,
RU's owning an equity participation
shown
to have
no
effect
if
to raise /2£/'s incentives
participation
C
venture and
Under
C
ship,
C
and
a good
is
RU
C
if
in
we do
C
has property rights over the innovation, and
has the property rights.
approximation when
first
C
ownership,
RU
process
It is
well
known (Binmore
we
will
theoretically arbitrary in that
is
posited;
%
Because
we
also
and 50
%
Under
RU
owner-
game (Rubinstein
(1981)) that this bargaining
assume. Of course, the exact
it is
split of
of-
the
sensitive to the bargaining process
would expect a variety of
is
splits in practice.
But we
that the innovator receives
of the pie (Caves et al (1983), Barton et al (1988)).
utility in the
evant range (in which
split
a small high-tech
when the time period between
should note that the industry rule of thumb
between 25
is
the standard alternating-offers
is
converges to zero, which
that
RU
freely uses the innovation.
process yields a licensing fee equal to V/2
is
Ignoring this equity
bargain over the licensing fee after the innovation appears.
(1982), Stahl (1972)).
pie
This equity participation can be
.
a large pharmaceutical or computer company.]
The bargaining
fers
not, but easily could allow for
C
ex post bargaining
game
is
transferable in the rel-
pays for the innovation), focusing on the equal
case involves no loss in insights, in that the qualitative adjustments are
straightforward for different
splits.
In contrast, utility
is
not always trans-
ferable in the relevant range in the ex ante bargaining over the allocation of
property rights and sharing
like to
(pV
C
—
rule.
For,
it
may be
the case that
RU
would
obtain property rights from C, resulting in a higher total surplus
e
— E) and
because
it is
that
RU
is
unable to make the compensating transfer to
cash constrained. Results
may
then qualitatively depend on
the division of the ex ante bargaining power. All the qualitative conclusions
may however be
obtained by looking, as we
10
will do, at
the two polar cases in
which one of the parties makes a take- or- leave- it contract
has
Remark By
focusing on ownership of the innovation,
:
possibility of
RU.
We
force
C
will
owning C. Suppose thus that
then assume that ownership of
C
to extract
V/2 from
C
a division owned by
in the
RU by threatening not
C
is
we have ignored the
does not imply that
produce and that C's indispensibility
to
it
RU
gains nothing from owning
to
RU
can
production process
produce
14
.
RU then
instead of only the innovation, and our focus
on the ownership of innovation involves no
is
and therefore
the bargaining power ex ante.
all
allows
offer
straightforwardly generalized
when
RU
loss of generality.
can force
C
[Our analysis
to produce.
We
will
not develop the analysis under this alternative assumption for the sake of
brevity.]
Optimal property rights when
in RU.
2.3
Suppose
C
Under
fore e
a=
in a first step that
=
0.
maximizes
ownership,
On
[p(0,
RU
(C has no equity
receives
the other hand,
E)V —
1
C
C
no reward
in
holds no equity
RU).
for innovating
15
and
there-
has appropriate incentives to invest.
E] and therefore chooses
E =
E*(V).
It
Utilities are
14
This is in the spirit of Hart-Moore (1991). Our assumption may be weaker than the
one made in Hart-Moore where parties are bound by a complete contract. Here, unlike
in Hart-Moore, the nature of production is ex ante ill-defined and cannot be described in
a contract. Note also that we do not make a similar assumption for RU (although we
could).
is
We
rather assume that
RU
cannot promise (contract on) the innovation before
it
made.
15
To be certain, there exists some noncontractual or informal sharing in that firms
reward successful researchers ex post through salary increases, cash awards, fringe benefits,
stocks or promotions. Such rewards are generally not commensurate with the value of the
innovation (Neumeyer (1971, chapter 3)). They play a slightly larger role for government
employees, perhaps because the employer's profit motive is lower (Neumeyer (1971, chapter
5)).
As we mentioned
earlier, e
=
is
a normalization for the
absence of formal rewards.
11
level of effort
exerted in the
sum
then (possibly up to a lump
transfer
C/R£/
C
from
RU):
to
=
(1)
and
Uc =p(0,E*(V))V-E'(V).
Under
mizes
[p(e,
RU
E)^ —
and so chooses
e]
E
its
investment so as to maximize
contractible and
C
C
Note that underinvestment by
RU
e
=
e*(-j ). If either
cannot ex ante compensate
£ is
E
is
maxi-
noncontractible
resulting in
ZJ],
occurs even when
C for an
sum
then (possibly up to a lump
£)y —
[p(e,
RU
in case of innovation.
has the bargaining power ex ante 16
or
is
¥
ownership, each receives
C
chooses
E =
E*(j).
,
contractible, because
increase in investment. Utilities are
transfer
from
C
to
RU):
V
V V
V
URU= P (e'(-),E'(j))--e-(-)
(2)
and
V
V V
V
V
Uc = p{e'{-),E'{^)) --E'C-).
How
are property rights determined?
having the property
16
If
E
is
contractible
of investment
E
rights.
and
RU
If
0,
RU
chooses
important enough that
has the bargaining power ex ante,
£
so as to
and so investment is socially optimal: E
If p(e*(^),£*(V))y < E'(V), then
optimal investment, which
is
consistent with C's participation,
C
to
RU
RU
Uc >
demand a
will
such that
prefers
C
level
breaks even:
+ E = p(e'(j),E)^.
P(e*(|).^)y " e'(y) +
cially
effort is
together with a cash transfer a from
a
Assuming a >
RU's
RU
Uru > Uru-
Clearly
maximize
"
=
=
p(e*(y),i?)V
- e*(|) -
E,
E'(V).
RU
should compensate
impossible.
and no cash
12
RU
C
for choosing the so-
then demands the highest investment
transfer.
Uc 17 then property
RU.
rights are allocated to
,
If
Uc < Uc,
the allocation
of property rights depends on the ex ante relative bargaining strength.
If
RU
is
has the bargaining power ex ante, the allocation of property rights
efficient in that
For,
C
If
RU
if
RU
receives ownership
ownership
ownership
is
RU
and only
if
Uru
+ Uc > Uru +
allocates the property rights to
gives the property rights to
In contrast,
keeps property rights as
Uru + Uc, an
RU
efficient,
of a cash transfer.
RU
efficient,
is
if
is
if
C
So,
C
when Uru
property rights occurs.
We
itself.
exchange
in
has the bargaining power,
cash constrained.
inefficient allocation of
C
Uc-
always
-f
Uc >
have thus
vindicated our earlier claim that ex ante bargaining power influences not only
the distribution of the pie, but also
Proposition
1
The
its size
18
allocation of the property rights
tween a research unit and a customer
(i)
is
on an innovation be-
determined by two factors:
underinvestment by both parties. Property rights are allocated to
the research unit
enough
(ii)
.
when the marginal
efficiency of its effort
is
large
relative to that of the customer's investment.
ex ante bargaining power of the two parties.
property rights
is
always
efficient
when the
The
allocation of
research unit has the
bargaining power ex ante, while the research unit's cash constraint
may
induce the customer to inefficiently retain ownership when
having the bargaining power ex ante.
17
For example,
+ t) v *
{'t
i.e.
18
if
and V*d
q(e)
=
in
2Xy/e and r(E)
= (t +
the marginal efficiency of
The
used
let
1?)
RU's
V2
-
= 2hVE.
Then Uc = H 2 V*,Uc =
C P refere RU ownerehi P if and only if A >
effort sufficiently
^/ft,
exceeds that of C'e investment.
idea that cash constraints reduce the efficiency of bargaining processes has been
the
field of
corporate finance (see Aghion-Bolton (1992)).
13
The
2.4
Suppose that
fee paid
=
a)£
by
irrelevance of C's equity in
C
is
given
C when
—
a) (nonvoting) shares in
RU. The
they agree on a nominal license fee £
The two
a£.
(1
RU.
parties then bargain over
is
then £
and nothing
£,
real license
is
= £—
(1
affected
—
by
the introduction of an equity participation (the reader will check this more
formally using the alternating-offers bargaining model). As explained in the
introduction, the customer
research unit
shows that
is
makes
RWs
it
is
handicapped by the
fact that its share in the
softer in the bargaining process.
C
taking a share in
A
similar reasoning
has no effect on the net license fee and
therefore irrelevant.
Remark
1
:
The
irrelevance result also holds for nonlinear sharing rules as
long as the two parties can by mutual consent tear up the initial contract
and write a new one. The
and not by the sharing
offer
bargaining
game
real license fee
rule. It
again driven by time impatience
can furthermore be shown in the alternating-
that forcing one party (here the customer, since the
research unit has no cash) to put
the other until "some agreement
for
is
up an hostage and pay a flow payment
is
to
reached" has no effect on the incentives
R&D.
Remark
in the
2
:
We
do not wish to imply that customers should never take equity
independent research units they sponsor.
does not raise the customer's investment since
transfer price.
But
it
it
Equity participation here
has no effect on the real
could affect other moral hazard components of the cus-
tomer's activity. For instance, in the presence of alternative customers (see
subsection 5.1)
could
resell
,
if
there were appropriability problems so that the customer
the technology to other customers, an equity participation in
the research unit would mitigate the customer's incentive to expropriate the
14
research unit (Rodriguez (1992)).
It
would
also soften future competition
between the research unit and the customer (see section 6 below).
Schumpeterian hypotheses.
2.5
The second most
tested set of hypotheses in industrial organization (after
the cross sectional analysis of the structure-conduct-performance paradigm)
relates
R&D
(R&D
input
expenditures or personnel engaged in
R&D)
output (as measured by the number of "significant" innovations) to
ables that presumably alter the incentives for
somewhat vaguely that a
"larger" firm has
interpretation of this scale effect
fits
from a process innovation
explanatory variable
of the firm.
An
is
is
R&D. The
more
or
vari-
scale effect states
incentives for
R&D. A
first
that a larger market for a good that bene-
raises the value of the innovation.
The
relevant
then the size of the business unit rather than the size
alternative view takes the size of the firm as the relevant
empirical variable. Relatedly, the scope effect (Nelson (1959)) posits that a
more
diversified firm exploits innovations
more
easily
than a specialized firm
and "therefore" has more incentives to innovate. The market power
effect
presumes that firms with market power gain more from an innovation.
How
do these hypotheses
in
fit
scope and market power
19
V.
effect all boil
down
Providmg general conditions linking these
tangential to the paper, but
it
our framework?
it is
seems natural to identify the
An
Let us assume that the scale,
to a change in
effects
V 19
and the value of the innovation
An
is
useful to give a few illustrative examples. For instance,
first
interpretation of the scale effect with an increase in
innovation that reduces the customer's marginal cost of production
the higher the output.
.
increase in the firm's scope
may
is
valued more,
also be viewed under
some
suppose that the customer may
produce two goods, A and B, thai may benefit from the innovation. Starting a product
line involves a fixed cost K. The innovation benefits either product line A or product line
B with equal probabilities and increases profit by V on the relevant product line. A firm
(strong) conditions as an increase in V.
that already produces
product
line
A
A
and
B
[To
has value
only values the innovation
illustrate this,
\V + ^ V = V, whereas a firm that has entered
jV + 5 (V — K) < V (assuming V > K). On
the other hand, one can think of circumstances under which a larger scope reduces the
incentive to innovate. Like in the case of a single good, an established firm using an old
15
First,
R&D
one can study how
inputs and output react to an increase in
the value of innovation, taking the organizational form as given (in the case
of cofinancing, the investor's share in the research unit
easily
shown that
is
kept fixed).
investment and probability of discovery
effort,
all
It is
increase
with V.
Second, we ought to ask how the organizational form (C ownership,
ownership with or without cofinancing)
for the innovation.
To
see
is
why a change
affected
in
V
by a change
RU
in C's value
has ambiguous implications,
V
let
us suppose that the customer has the ex ante bargaining power and ignore,
examples we
as in this section, cofinancing (in our
that cofinancing
is
indifferent
is
V
aRUv
be preferred by C, since
>
0.
technology
ownership.
The second
may have
from scratch
so
such that the customer
C
E RU = p Cy _ £C
V A first
.
is
e '(^),E'(^))
effect is that
if
p
m
=
p
c
replacement
and p c
C ownership
=
C
p,
goes in the other direction:
less incentives to discover
(this is the familiar
(
residual claimant under
For instance,
effect
and aCv to designate the owner
satisfy:
E™ = E\\),E C = E'{V),p m = p
Consider a small increase in
p/2
must
RuY_ _
p
RU
make assumptions
between keeping property rights and allocating them to the
of the innovation, such a
under
V
indeed suboptimal). Consider a
research unit. Using the superscripts
where
will
A
should
now
ownership and not
£?{p
cV
The
a new technology than
effect).]
= p(0,£*(V)).
- p^j) =
research unit's
if it
were starting
variation of the scope
argument
the idea of "absorptive capacity" (Cohen-Levinthal (1990)): It has been argued that a
firm's ability to recognize the value of external innovation and assimilate it properly is a
is
function of the firm's level of prior related knowledge generated in particular by
R&D effort.
its
own
market power effect can also in some circumstances be formalized
by an increase in V If either an increase in market power raises the output produced (to
the detriment of product market rivals) or the innovation cannot be perfectly protected
by a patent, trade secret or internal use and therefore spills over to rivals, an increase in
market power may well raise the firm's valuation for the innovation.
Lastly, the
.
16
ership.
that p
c
There
V
with
effort increases
is
under
RU
ownership but
is
unaffected under
C
own-
a third and more ambiguous effect resulting from the fact
in general differs
from p RU
But the
.
first
two
one cannot draw any general conclusion about the
show why
effects clearly
effect of scale,
scope or
market power on the management of innovation. When innovating becomes
more important,
customer wants
the
more of
to appropriate
it
ownership, but also to raise the research unit's incentives through
ership.
in
Appendix
1
illustrates this
Consider a switch from
C
available input measure, the customer's
ownership
for
to
RU
be measured 20
measure and the value
The
declines.
may
well
Input and output
opposite directions because part of the input cannot
Furthermore, there
.
re-
The
ownership.
monetary investment,
increase due to improved incentives of the research unit.
in
own-
Schumpeterian
output measure, here the probability of discovery, on the other hand
measures then move
RU
with two simple examples. The changes
ownership structure have important consequences
lationships.
C
through
is
no clear relationship between either
of the innovation
and therefore the Schumpeterian
parameters.
Remark The
:
analysis of section 3 will point at
culty with empirical
tional form, as
work using
R&D
still
another theoretical
input or output, but not the organiza-
dependant variables (besides the traditional
the measurement of variables: see Cohen-Levin (1989) for a
sion of these issues): Input measures
commitment
we assume
may
20
to
R&D
may
also
choose to alleviate
its
See example 2 in Appendix
is
issues concerning
summary
discus-
be misleading because a firm's
cannot be simply measured by
that this effort
diffi-
purely monetary.
its
In our
R&D
effort
even
model the customer
investment burden by demanding cofinancing.
1
for
an
illustration. Fisher
empirical discussion of input and output measures of
hypothesis of increasing returns within
R k. D
R&D and the scale
17
if
Its
and Temin (1973) provide an
activity, focusing on the joint
effect.
own
R&D
vestor)
R&D
much
too
expenditures
may
decrease while consolidated (customer plus
expenditures increase.
significance to firm level
It
in-
therefore seems natural not to attach
R&D
input viewed in isolation.
Multiple financing.
3
The
3.1
rationale for cofinanding.
Parties that do not directly benefit from the innovation,
ture capitalists or a parent
research unit.
To explain
company
this,
namely banks, ven-
often contribute to the financing of the
we must come
to grips with the issue of
what
why
can't
such investors do that customers could not do themselves; that
customers take on the share held by investors?
to extract the research unit's rent provides
As observed
not dilute
in section 2.4,
RVs
too. In contrast,
income (under
We
argue that the desire
a rationale for this institution.
RU
C's acquisition of a stake in
return from innovation:
C
ownership).
We
actually does
cannot have the cake and eat
suppose that an investor claims a fraction
RU
is,
keep assuming that
RU
(1
—
and
it
a) of RU's
C
bargain
over the licensing fee after the innovation occurs. Even though the research
unit receives only a fraction
and
C
is
unaffected.
RU
a
then receives a return
compared with V/2 when the
of the investors.
of the fee, the bargaining process
Note that
(1
—a)
qV/2
is
shares are held by the customer instead
in the bargaining process the investors
possible; therefore the investors
The
lack of incentive to collude
comes from the
and the
much from
the
have no incentive to enter the
bargaining process or to collude with the research unit
21
RU
for the innovation as
research unit have congruent interests, namely to extract as
customer as
between
21
.
linear sharing rule.
The corporate
finance literature has demonstrated that, in the absence of renegotiation or collusion,
a nonlinear contract between an investor and an agent, such as a debt contract, can
strengthen the agent's bargaining position with a third party. Nonlinear contracts on the
18
Remark That
:
the research unit's return from innovating
(customer, investor) owns
investor
is
drawn
its
by who
affected
is
shares carries over to the situation where the
To
into the bargaining process.
illustrate this,
suppose
that the three parties bargain according to (a three-party extension of) the
Each period, one of the three parties
alternating-offer model:
equal probabilities, to
make an
offer to the
(resp. the research unit) accepts
research unit can strike a deal between
agreement; the investor
is
two parties while
customer) yields an agreement
offer of the research unit (resp. the
customer
drawn, with
other two. Let the initial contract
specify that the investor needs the agreement of the other
an
is
it.
That
is,
its
share
the
the customer and the
them without needing the
protected only by
if
(1
— a)
in
investor's
whatever the
research unit receives. In a stationary equilibrium of this bargaining game,
the research unit's payoff
factor between offers.
as the offers
It
is
equal to aV/(3
—
thus converges to the
become very
frequent.
The
8),
where S
same payoff
is
the discount
as above (aV/2)
payoffs of the investor
customer also converge to those obtained when the investor
is
and the
not part of the
bargaining process. To summarize, under alternating-offer bargaining,
initial
property rights can be chosen so as to yield the same sharing rule as when
only the customer and the research unit bargain.
Our
analysis studies the best case for cofinancing,
namely when customer
and investor investments are contractible and perfect substitutes.
When RU
has the bargaining power ex ante and C's participation constraint
binding in the absence of cash transfer, then
RU
is
not
does not offer cofinancing
other hand are sensitive to the possibility of secret renegotiation between the investor and
the agent. Indeed
if
the research unit and the customer bargain together,
it is
optimal for
the research unit and the investor to secretly renegotiate towards a linear sharing rule so
as to obtain congruence.
19
because
would dilute
this
its
for a given level of financing
share and lower
to investors yields
incentives. If
however
22
to an investor
When C
of
cash than
it
—
a) of
from reduced
loses
R&D
even in the absence
to alleviate C's financing
burden by resorting
has the bargaining power ex ante and
RU ownership
Ei from a competitive investor
RU's
for a fraction (1
in
is
i?£/'s incentives
optimal for C,
C demands
are
suffi-
cofinanc-
exchange of a claim of a fraction
(1
—a)
where
profit,
EI = p(e'(aj),EI + Ec)(l-a)j,
'
and where we ignore any transfer aj from the investor to the customer
the moment. C's profit
E = Ec + Ej is total
customer
22
is
is
Formally,
E—
let
E = Ec + Ei),
E'((l
—
for
then
is
p(e>|), £)| - Ec = P(e>y), E){\ - |)V where
is,
.
ciently important that
ing
less in
does not want to invest in
RU may want
of cash transfer,
RU
That
research effort.
by C, offering a claim
RU's income
C
its
investment.
The optimal
E,
total investment for the
y)V). Cofinancing thus allows the customer to give
Ec and Ej denote customer and investor investments (total investment
and ac > and a/ > denote their cash payments to RU. The research
unit must guarantee that the customer and the investor get their individual rationality
levels:
Ec +ac<p(e'(*^),E)j
and
Ei+a,<p(e'(aj),E)(l- a )^.
The Lagrangian
C=
maximization of RU's
p— -e + ac + ai + n
So either
||
for the
(e*
=
ft
A
=
1,
(a£) E) (2f ) =
,
The Lagrangian then may
e is inelastic in
Py - Ec - ac + A
and then
1.
So,
a =
a 6
utility over
P(!
V
"
»)
[0, 1] is:
J -Ei—aj
+ 7<*c
+£<*!
any E, this Lagrangian increases with a since
and a/ = £, = 0. Or /i = A > 1 and a { = a c = 0.
for
1
decrease with
a
at
a =
1.
This
is
in particular the case
when
o, so cofinancing relaxes the customer's participation constraint without
worsening incentives.
20
the research unit any fraction of the value of the innovation between
ownership) and 1/2 (pure
RU ownership).
(C
With contractible, perfectly substi-
tutable investments, cofinancing transforms a discrete choice of governance
structure into a continuous one.
The
a
choice of
is
based on two considerations.
First, a lower
a
allows
cofinancing and thus reduces the customer's investment burden.Second, the
dilution of /?£/'s share reduces
Using
discovery.
RWs
its
first-order condition (tfie'l 3^)) 2^-
tive of C's profit with respect to
term
in this derivative
That a
centive effect.
is
a
is
equal to [—
lution (that
is
is, if
=
is
The amount
Ej
Remark
1
=
1)
E] < E*((l
is
is
a=
large), then
1 is
or
is
may
The
first
is
not
not be optimal
quite sensitive to di-
indeed optimal. But
if
effort
dominates and cofinancing
of investor financing for the optimal a'
If this
is
inequality
is
is
then given by
optimal as long as
not satisfied, one optimal policy for
to let the investor finance the whole investment
and ask
for
a transfer a/
Remark
2
When
=
E m ((l — ^)V)
E\ — E'{(\ - fyV).
customer and investor investments are contractible and
perfect substitutes, as was
presumed
The
research unit's incentives under
ting
a =
0,
^r)]-
= P(e>*|), JET((1 - y)V))(l - a*)y.
— %-)V).
the customer
:
may
Not requiring any transfer from the investor
:
the deriva-
never optimal implies that (7-ownership
relatively inelastic, the rent extraction effect
occurs.
1),
V + ^r(~
size of the incentive effect. If effort
—q"/q'
=
the rent extraction effect and the second the in-
optimal either. Pure customer financing (a
depending on the
and reduces the probability of
incentives
and the
C
here,
C
ownership
is
never optimal.
ownership can be duplicated by
investor's benefit can
be used either to reduce the
customer's investment burden or as a cash transfer to the customer.
21
set-
Fur-
thermore, increasing a slightly above
under
the customer's welfare unless the effort supply
C
RU
is
ownership
strictly raises
completely inelastic.
ownership can be optimal when the customer sinks noncontractible
investment such as advice on design or release of proprietary technological
information.
C ownership
Then
cofinancing at
a =
tion function.
A move
and
if
cannot be duplicated by
Ej and Ec are not perfect substitutes
to
RU
We
financing.
in the
with
produc-
ownership reduces the customer's investment
investment cannot be perfectly
this
RU ownership
offset
by an increase
in the investor's
conclude that customer ownership can result from the exis-
tence of noncontractible investment by the customer.
Remark
3
While the model predicts that cofinancing can occur,
:
distinguish
among
To
types of investors.
it
does not
predict whether the investor should
be a bank, a venture capitalist or a parent company requires a richer framework.
may
For instance,
it
may be
that a research firm (as a parent
company)
help the research unit through the noncontractual release of techno-
Or a parent company may have better information than the
logical advice.
financial
market about
to invest in
may be
it.
this
type of technology and therefore be more willing
Conversely, independent financing by financial intermediaries
desirable in the absence of a perfect capital market.
Proposition 2
(i)
.
Suppose the customer's and the
tractible
and perfect substitutes.
investor's investments are con-
Customer ownership
is
never
optimal. Ownership by the research unit involves no cofinancing
by an investor
if
either the research unit has ex ante bargaining
power and the customer
is
willing to participate or the research
unit's effort decreases rapidly with a dilution of its equity. Other-
22
wise cofinancing can occur.
Customer ownership can be rationalized by the existence
(ii)
of non-
contractible investment by the customer.
3.2
The long-purse
effect.
A
This subsection studies another aspect of the financing of R&D.
Schumpeterian hypothesis
its
ability to raise funds.
relates a firm's
It
well-known
R&D investment to its size through
has been argued that being large
is
positively
correlated with having a long purse, and that, in an imperfect capital market,
a long purse permits more investment, in particular
That a
R&D
when
firm's wealth lowers the cost of investment
incomplete information about
its activities
Suppose
creditors have
has been well established by the
corporate finance literature. But the long-purse effect
similar to those in subsection 2.5.
investment.
is
subject to caveats
C
for instance that
ownership
is
optimal when the customer can finance E*(V) without going on the capital
market.
Consider now the situation in which the customer has
E'(V) and must borrow. The management
by the need to borrow
First,
RU
in
of the innovation
may be
ownership becomes relatively more attractive since
(
yj < E* (V).
RU
actually increase the probability of innovation.
in
it
requires
A move
from
ownership or cofinancing indeed lowers the customer's
R&D investment, as predicted by the long purse
an example where
affected
Second, cofinancing also
allows a reduction in the investment burden of the customer.
ownership to
than
an imperfect capital market.
a lower monetary investment E'
C
less
effect.
Appendix
But
this
move may
2 illustrates this in
the absence of credit-constraints, i2£/-ownership
is
dom-
inated by C-ownership whereas the opposite holds once credit-constraints are
introduced.
23
Another conclusion
is
the initial distribution of assets influence
locations of property rights on innovations.
new
activity than established, healthy firms.
that monetary
R&D
al-
we would expect
firms or firms which have experienced hard times to farm out
R&D
is
In particular
the
more
their
In this sense the prediction
investment should be positively correlated with assets
not invalidated by our analysis.
Property rights and the drasticity of
4
in-
novations.
This section investigates the relationship between property rights and the
nature of the innovation: Does an independent research unit have more
in-
centives to pursue radical (drastic) vs incremental (nondrastic) innovations
compared with the same research unit within an integrated firm?
"Drasticity" here refers to the size of the technological or quality improve-
ment brought about by the innovation. For example, the
cost reduction, the
single innovation
more
drastic the process innovation
larger the marginal
is
23
We
.
and assume that the probability of discovery decreases with
the size of the innovation. To focus on the choice of technology,
the inputs e and
E in
The parameter 7 >
line",
with 7
1
we
ignore
the notation and denote the probability of discovery by
p(~f).
=
consider a
1
indexes the "size of the innovation" or "research
corresponding to the existing technology.
ously said about inputs
still
applies
What was
previ-
and of course influences the distribution
of property rights.
Following the patent race literature,
we
refer to the current
customer
C as
^See Aghion-Howitt (1992) for a general equilibrium model of vertical innovations where
the size of quality improvements is also treated as a choice variable available to research
firms.
are
However, as
in the
whole literature on
left aside.
24
R&D
and patent
races,
ownership aspects
the "incumbent" and to other potential customers as "entrants". Let ir™(f)
denote the incumbent's profit when he obtains an exclusive license
innovation, and
let 7r™
=
purchased an exclusive
if
the innovation
technology
is
denote the incumbent's (monopoly) profit in
71-^(1)
the absence of innovation;
for the
finally, 71-1(7)
One has
license.
*s
the profit of an entrant
71-1(7)
5:
who
has
n?(l)i with equality only
"drastic", in the sense that the entrant
owning the new
not constrained by the competitive pressure of the incumbent.
is
Since 7 indexes the size of the innovation, 7r™(-) and ^i(-) are both increasing.
We now
C
compare the
RU)
(respectively
size.
[We
sizes
7c and ^ru of the innovation that obtain when
has property rights on the innovation and chooses
will also allow
property rights to be
split
its
from control rights over
the process.]
Under
C
ownership, we have:
7c
Note that 7c
is
=
arg
maxp(7) [7^(7) -
tt™]
.
By
the efficient research line for the industry.
a revealed
preference argument, the larger 7r™, the more drastic the innovation 24 This
.
is
a version of Arrow's (1962) celebrated replacement
effect.
The incumbent
prefers a lower probability - higher payoff research technology
the absence of innovation
is
if its
profit in
high.
Note that the same choice fc would be made under C-ownership with
the research unit controlling the research
fine, since in
bargaining power and no payoff ex post and therefore
So there
is
no gain
Consider now
in splitting
RU
that case
is
RU
has no
willing to choose -yc-
property rights and control rights.
ownership (with or without cofinancing). Note that in
the absence of potential entrants, the research unit gets half of the value of
34
As
usual, this
is
a set comparison
if
there are several optima.
25
the innovation in either case and therefore chooses the technology preferred
by C: fmj
=
7c- This congruence between the research unit's and the incum-
when the
bent's preferences disappears
Although
research unit can
in equilibrium the research unit
optimally
to the incumbent, an entrant acts as a threat
sells
sell
to
an entrant.
an exclusive license
and creates an appropriation
effect.
More
precisely,
if
the innovation
licensed by
is
RU
to a potential entrant
(instead of being licensed to the incumbent), the entrant can generate a profit
equal to 7^(7). In other words, the research unit has the outside option to
an exclusive license to an entrant
sell
in
Binmore
et al (1986)
at price ^"1(7).
A
well-known result
then shows that, under alternating-offers bargaining
with this outside option, the research unit can obtain a license fee equal to
max ( 'i"^"""
Suppose
,
7^(7)) from C.
first
that the optimal choice 7/^7 by the research unit
is
such
that:
>
2
Then, the outside option of
selling to
mimh
an entrant
is
irrelevant
and
therefore,
as in the absence of a potential entrant, the research unit chooses the
technology (or technologies
customer: imj
On
Iru
=
-y
C
if
same
there are multiple optima) as the incumbent
-
the other hand,
will generally differ
if
K\{iru)
>
*'
nS£J-Ia
)
the research unit's choice
from the incumbent's 7c since now the outside option
of selling to an entrant
becomes credible and the incumbent must pay ir\(lBu)
to obtain the license.
More
m
P(7c)kr(7c)-*o
]
precisely,
by definition of fjuj and 7c, we have:
> p(lmj)[*?(fRu)-*Z]
and
26
Multiplying these two inequalities, we obtain:
In particular,
we have
Proposition
3: Interests
when
established the following proposition:
may
diverge as to the choice of the research line
Let r(7)
there are potential entrants.
= ^J'r^
denote the
appropriability ratio or relative willingness to pay of an entrant w.r.t.
the incumbent for the innovation. Then:
In other words,
if
Ibu > 7c
if
r is strictly
increasing,
Iru < lc
if
r is strictly
decreasing.
and
the relative willingness to pay of a potential entrant in-
creases with the size of the innovation, then the research unit will tend to
choose a more drastic innovation than the incumbent, since by doing so
it is
able to appropriate a higher fraction of the surplus of the incumbent. Similarly, if
the relative willingness to pay of a potential entrant decreases with
the size of the innovation, the research unit appropriates
plus by choosing a less drastic innovation
interval
and
25
.
strictly decreasing over another,
If
r
is
then
more
of the sur-
strictly increasing over
it is
possible to find two re-
search technologies p(-) and p(-) such that (with obvious notation) fm;
and
25
7^
>
an
<
jc
7c.
The patent
race literature (Gilbert-Newbery (1982),
that an incumbent
is
more
likely to
Reinganum (1983)) has stressed
when the innovation is
innovate before an entrant
the innovation is drastic. On the one hand, the replacement
incumbent is not in a hurry to innovate; on the other hand, competition destroys industry profit and therefore the incumbent gains more from remaining
a monopoly than an entrant from becoming a duopolist. The latter efficiency effect is
absent here, since the research unit sells to the incumbent anyway. Our analysis, which
also hinges on the willingness to innovate of the incumbent and the entrant, has a very
different focus, namely the choice of research technology for a single research unit rather
than the race between two research units with fixed research technologies.
minor, and conversely
when
effect implies that the
27
What about
split
impose the research
RU. Then
=
7*
C
Clearly,
C
property and control rights?
line,
chooses to maximize his expected net gain from the innovation:
arg
max
|p( 7 )
min
|
^^
^(7) - *« - ^(7)) }
,
>
'
is
~
.
would never choose a drastic innovation, since he would prefer
—
v T(l)
able to
is
but that ownership of the innovation remains with
"not to innovate (choose 7 = 1 and obtain 0). If '
2
T m(7 T °m
j~
7* = 7c- If
< 7^(7), then noticing that 7^(7) -
(
C
Suppose
—
1
^o
)
(1
r (7))> 7*
strictly increasing.
— 1c
if
r
strictly decreasing
is
tt
^1(7), then
™ -
=
7^(7)
and 7* < 7c
if
r
Quite naturally, we find that the incumbent prefers to
reduce the research unit's bargaining power and thus imposes a bias in the
research orientation opposite to that desired by the research unit.
Example Process innovation
:
in
an homogenous good industry The incum:
bent monopolist produces at marginal cost
The
cq.
potential innovation
process innovation that reduces the marginal cost to c
an entrant purchases the exclusive
petition. Letting
license, the
<
cq.
So 7
=
is
a
Co/c. If
two firms wage Bertrand com-
D(p) denote the demand curve. One has (abusing notation
slightly)
*i(c)-i?MJ* bbT(r))dx
*x(c)
As
is
well
=
min {D
(co)
(a -
c)
,
D (p m (c)) (p m (c) - c)}
.
known, an entrant has more incentive to innovate than an incum-
bent due to the replacement
that the outside option
is
effect so r
1.
A
>
fortiori, Xi(c)
'
—
j
"-,
so
binding regardless of the size of the process inno-
vation. Furthermore, r decreases
(7 increases).
>
from
D (cq) /D (pm (cq))
to
1
as c decreases
In an homogenous good industry, an independent research
unit will pursue less drastic innovations than
28
an integrated one. Appendix 3
studies two other standard models of industry behavior, that yield ambiguous
conclusions as to the monotonicity of the appropriability ratio.
Remark
1
(exclusive dealing).
We have seen
that
when 7Ti (7/^) >
T
i
Hgglzfn
)
the research unit chooses an inefficient research line in order to increase
its
bargaining power. This inefficiency can be eliminated while leaving owner-
by signing an exclusivity contract.
ship
and control
The
outside option being removed, the research unit's interests would then
rights to the research unit
match the customer's. However,
trants raises
its
letting the research unit's trade with en-
payoff in case of innovation and thereby
exert effort. There
is
its
incentives to
a strong analogy here with the comparison between a
genuine M-form and a corrupted M-form (see HolmstromrTirole (1991)). In
a corrupted M-form, each division
is
prohibited from trading with outside
partners and chooses the socially optimal specialization of
In a genuine
its
investment.
M-form, divisions can trade with outside partners. They then
do not specialize
their investments in order to gain a bargaining power; but
the existence of competition gives them more incentives to invest than in a
corrupted M-form.
Remark
2 (antitrust law).
Suppose that an antitrust law prevents the
search unit from licensing to the incumbent.
pendent) then chooses fm; so as to maximize
between ~(ru and 7c (the
in Proposition 3.
units' profits
precluded 26
26
under
level
The
research unit
p(-))iT\ (7).
RU
ownership because
inde-
The comparison
chosen by an integrated firm)
However, fc no longer maximizes the
(if
re-
sum
is
as given
of the two
efficient licensing negotiation is
.
Letting To (7) denote the incumbent's profit
license, the joint incentive to innovate is [^1(7)
equality for a drastic innovation).
29
when an entrant obtains an
+
(fo(7)
—
^o*)]
^ *T(f) ~
exclusive
*<? (with
Split
5
We
property rights.
have analyzed the allocation of a single property
innovation
may have more than one customer
one innovation. There
is
A
or there
may be more
that
then more than one property right to allocate, and
property rights can, and often are
5.1
In practice the
right.
split.
rationale for shop rights.
This subsection derives foundations for the observed practice of employers
allocating ownership to employees while keeping a royalty-free, nonassignable
licence for themselves. Consider the following situation.
signed and
An
managed through a contract between a research
innovation deunit
RU
and a
customer C\ can ex post be sold to a second, yet unidentified customer C?
as well as to C\.
and
V2
C\ and
C2
do not compete on the product market;
denote their valuations of the innovation, where \\
E
non-contractible investment
for the innovation [e.g., V\
=
Vo
C2
with other firms
in its industry.
which
dustry,
C2
V2
reflect
some
by C\, that increases C\S willingness to pay
convention that
at
may
let V\
+
E, where Vq
a constant.]
is
purchases the innovation at price
More
.
adopt the
C2
competes
So
generally "Vj" could stand for the price
purchases the innovation
would be equal to half of
V2
We
C2
(so, if
C2
is
a monopsonist in
valuation in the
's
its in-
Nash bargaining
solution).
Let Vfaj and Vc, denote
RVs
the extra payoffs they obtain
and C\S stakes
when innovation
VRU + VCl
=V
1
VRU <-± + V2
30
occurs.
+ V2
and
.
in the innovation, that
One
has
is
The
latter inequality
RU
does not extract more
owns the innovation.
Because both parties
comes from the
than Vi/2 from C\ even
if it
fact that
never have an incentive to overinvest (their individual stakes never exceed
Vi
+ ^)>
an optimal contract must maximize the stake of either party
innovation given the other party's stake
of Vi
V When
and
2
there
.
innovation (E
=
0),
27
Let o^ and a? denote M/'s share
.
no customer investment
is
in the
an optimal contract must
in raising the value of
in particular solve:
max{(l-Q )y + (l-a 2 )V2 }
1
1
subject to
ai^i
+ a 2 V2 > Vpu
and
<«i < -
We
<a
,
2
<
1.
thus focus only on the allocation of incentives; the optimal V^v
mined by the same considerations
in section 2. In this
only on
RVs
program a\
total share (a\V\
is
on
to C\
words,
much
if
its
>
own use
C\ can
0),
indeterminate, as both incentives depend
+ ct 2 V2 ).
However, when Cj's investment does
of the innovation.
but not
affect Vi
27
V2
,
In particular
its
This
ment. It
is
is
RU
and
1
—
and
at\
=
if
if
V^p <
V2
,
a2 <
be
1.
In
tilted as
the optimal con-
allocates the licensing fees in proportions
a\ to C\.
We
thus obtain a rationale for shop
if both parties incur some (even negligible) non contractible investwhen the customer's whole investment is contractible that the customer's
the case
only
valuation V\ for
relative share should
as possible towards the first use. Indeed,
^rlt/^2 t°
own
its
the optimal contract must give maximal incentives
tract gives a shop right to C\
Q2 —
deter-
(incentives, ex ante bargaining powers) as
not only affect the probability of discovery, but also
the innovation (E
is
stake becomes irrelevant.
31
rights
5.2
28
.
Multiple innovations: contingent property rights.
We observed
introduction that both employment contracts and the law
in the
how much customer investment was
allocate property rights on the basis of
used by the research unit and of whether the research unit had been hired
and made
for the innovation
during normal working hours.
it
We
argue that
these contingent property rights stem from incentive considerations.
Coming back
the investment
With
to the single
E
can yield one
probability x
Some
relevant one.
customer case, suppose that the
t ,
in a subset
with ^2teT x t
=
1?
V
t
+ of types of innovation.
innovation
t,
with value
V
t
,
is
the
by alternative customers (so the "customer"
can be both a customer and an investor
also
3R
and
types of innovation are consumed by the customer, some
others are purchased at price
may
T C
effort e
be demanded by both, as
in
our terminology). Innovation
in subsection 3.1.
types of innovation differ in the extent they
t
Besides their value,
make use
of the customer's
investment. Namely, we assume that the probability of discovery conditional
on innovation
t
assume that
and
relevant 29
t.
.
e
Let
being the relevant one
E
q(e)
+
tr(E).
We
also
know which innovation
is
share of the value of the innovation of type
Note that the nature of innovation
we maximize C's
=
p(e,E,t)
are chosen before the parties
a denote RU's
t
is
is
contractible.
As
in subsection 3.1,
incentive subject to i?£/'s incentive exceeding
some
level:
max£, 6T z,*(l - a )V
t
28
When
>
t
RU
enough of the
must then be used.
^This is the simplest version of the model. One could alternatively assume that effort
and investment are contingent on feasible types of innovation.
Vjjy
Vj,
a shop right does not enable
to appropriate
innovation. Other contracts (possibly including cofinancing)
32
subject to
w
E teT x a V > V
t
t
t
and
<a <a
t
where
a =
t
a = 1/2
(as in subsection 3.1)
1 if
program
t
if
t ,
C
consumes the innovation and
an alternative customer uses the innovation. The solution to
satisfies for
tQ
=»
=>
=>
policy
is
t
>
<
t
=
t
The optimal
some
t
t
in the introduction.
this
£ T:
t
a =
a = a
<a <
(C
t
t
t
ownership)
RU
(pure
t
ownership)
(sharing).
citQ
thus very similar to the contingent splits described
If
the innovation makes
much
use (respectively,
little)
use of the customer's investment, the customer (respectively, the research
unit)
owns the innovation.
For example,
if
customers, this
innovation
split
may
customers use innovation
is
t
tQ
,
is
the optimal contract gives
required unless
to.
If
[If
a to =
or 1/2.
— 2a
rights but also the "hired for" doctrine.
make more use
is
therefore should be
owned by the employer.
is
tQ )
As shown
of
RU's
in section 3,
an
equity.]
split of
property
of the employer's investment
and
that our assumption of a single effort has swept aside a po-
tential inefficiency in the allocation of effort created
rights. If not
a share a to of
Presumably, the innovations for
which the employee
caveat
only alternative
RU
This analysis rationalizes not only the usual contingent
A
.
only the customer uses the innovation,
investor ought to purchase a fraction (1
hired
io
used both by the customer and alternative
take the form of a shop right.
the licensing fees on innovation
cofinancing
V
In the middle case, they split the benefit
by contingent property
monitored, the agent has an incentive to concentrate her atten-
33
tion
on those potential innovations with the highest a
raises the desirability of
uniform property rights
to noncontractible investment
distinctions
made between
30
.
t
.
Such
effort allocation
[A similar point applies
by the customer.] This caveat rationalizes the
research done at
home and
the employer can better monitor the effort allocation
at work.
Presumably
when the employee
is
at work.
Multiple innovations: sequential property rights.
5.3
Many employment
employee shortly
An
contracts specify that an innovation
after quitting the firm belong to the former
date.
To the
is
.
nature of the innovation by
extent that the current innovation has a higher current
employer investment content than the future one,
^Thifl reasoning
employer31
can be obtained by following the
efficiency rationale for this practice
lines of the previous subsection, replacing the
its
made by a breakaway
it
seems reasonable to
Holmstrom-Milgrom (1991)
and 8) for the case
on optimal taxation and Ramsey pricing.
familiar from multitask models. See, e.g.,
for the case of agent risk aversion, LafTont-Tirole (1993, chapters 3,4
of adverse selection, and the related literatures
31
This
is
so despite the facts that such contracts
of employees and that they
3) describes
some such
may be
may
inefficiently restrict the mobility
vulnerable to legal attacks.
"trailer clauses"
,
Neumeyer
(1971, chapter
for instance:
"At Polaroid, the provisions of the [assignment] contract are valid for all
made or acquired during employment or for one year after termination of employment" "The agreement is also equipped with a trailer
clause valid for two years after the termination of employment
The eminventions
.
.
ployee will not contribute his knowledge to
.
.
.
.
.
any corporation or person
engaged in competition with Polaroid."
At Gulf Oil, "[employees in technical or scientific work] agree ... for one year
after employment not to engage in the same type of work for a competitor
within the same geographical area or territory. This second part of the
[trailer]
clause covers practically
all
employees who are expected to make
patentable inventions and improvements."
Note that such clauses have the
efficiency properties described here (to the extent that the
content of the employer's investment in the post-employment innovation
the second innovation
is
in a
is
larger
competing area) or have anticompetitive motives.
34
when
allocate property rights in this
We
manner 32
.
summarize the section with the following proposition:
Proposition 4
:
In the presence of multiple users or multiple innovations,
the property rights
unit.
may be
between the customer and the research
split
Each should get property
on those
rights
activities for
which
it
has a comparative advantage in creating value. This principle gives rise
to shop rights, property rights based
on the nature or the date of the
innovation and to the "hired for" doctrine.
The dynamic management
6
of innovation:
short-lived vs integrated research ventures.
Nowadays, large american companies such as
hundreds of research joint ventures (RJV).
as "an association of two or
owners a
A joint
or
AT&T
venture
is
more natural or juridic persons
specific project for a limited time"
purpose of
IBM
this organizational
form
is
to
32
is
to
The
The second innovation
to carry
on as
co-
promote an entrepreneurial and non-
The
short-lived
be contrasted with more durable and
may
Example: Suppose that the innovation, with value V\,
innovation with value Vj.
typically defined
(Brodley (1982)). The declared
bureaucratic spirit in research and development activities.
nature of the research joint venture
are involved in
probability of the
first
give rise to a follow-up
innovation occurring
is q\
(e\)
+
RU
and has probability q?{ei) of
occurring conditionally on the first innovation being made. Letting ax an ^ a 2 denote
RU'b share of the values, we can as in subsection 5.2 maximize the customer's payoff from
the first innovation, (1 — a x ) V\ + p? (ej (<*2 Vj)) (1 — 02) Vj fixing RU'b payoff from that
innovation, 0:1 Vi + [p2 (e\ (0:2^2)) 02^2 — e\ (otjVa)], and given the constraints
< o« < d«,
where a t = 1/2 or 1 depending on whether C is the user of the innovation. We will assume
ri (Ei).
requires only effort by
,
that
>
—
C
does not use the second innovation: d 2
=
1.
Solving this program,
we obtain
> and
owns the first innovation and shares the benefit of the second, which may take
the form of a minimal length of time over which the innovation does not belong to RU.
[If C's investment content in the second innovation (which here is assumed equal to zero)
declined over time, the use of such a minimal length of time would be strictly optimal.]
Or a 2 = 1 and at\ < 1: RU claims part of the first innovation and is residual claimant for
02
a\
01. Indeed,
0:
if
positive incentives are given to the
C
the second.
35
two
parties,
then either 02
integrated relationships created by (nontradable) direct equity participations
or full vertical integration. Not surprisingly, the appropriability of (specific)
knowledge
a major consideration in the choice between these various forms
is
of collaborative ventures
33
.
The purpose
of this section
is
to formalize (spell
out) the basic trade-offs underlying this choice.
As
before,
we consider the
relationship between a research unit
a development unit (user or manufacturer) C.
its
The
and
research unit contributes
knowledge and the manufacturer contributes capital or other inputs to the
production of a current innovation aimed at a third party
of the research unit's knowledge cannot be contracted upon.
transfer
from
from
i?£/'s
What
is
RU
RU
to
C
perspective
is
will thus
only take place
if it is
34
.
The
The technology
incentive compatible
35
.
the rationale for such technology transfers?
The primary
goal
More
pre-
to increase the probability of the current innovation occurring.
cisely,
33
An
we assume
transfer
that the probability of
making the purported innovation
interesting study by Pisano-Russo-Teece (1988), based
on data from 974
collab-
orative arrangements in the U.S. telecommunications sector over the period 1982-1985,
analyzes the choice of governance structure in the case of an
R&D firm collaborating with
R&D firms engaging
a manufacturing firm (as opposed to the "horizontal" case of several
R&D services for the development of new
Brodley (1982) often consider research joint ventures as a
in cross-licensing or in the provision of joint
products).
Legal experts
like
special case of "input joint ventures".
^That a
third party rather than
C
uses the innovation allows us to cleanly identify
the dynamic considerations behind the choice of ownership, since from a static viewpoint
incentives are entirely provided by the rule for sharing the income from this third party and
ownership is irrelevant. One could clearly reintroduce the static dimension of ownership
by assuming, as in sections 2-3, that C is the only user of the innovation. However, this
would only restrict the set of feasible sharing rules over the first innovation without adding
any insight as to the basic trade-offs governing the dynamic management of innovations.
^For example, in the biotechnology industry, "technical and operating procedures are
not well codified or even understood" .To partly circumvent the resulting difficulties for
technology transfers, contracts do often provide for a high degree of contact and exchanges
between the parties (e.g. by including a schedule of telephone conferences and joint meetings to aid in the transfer of information, in addition to stipulating the exchange of written
documents." (Pisano-Mang (1992)).
36
is:
qQ
P=
+
r(E)
RU
transfers its technology
if
RU
does not transfer
\
r(E)
denotes the manufacturer's investment 36
E
where
if
noncontractible (alternatively
E
incentive effect" described below
The innovation
rule (cm for
in the
RU,
1
is
its
technology,
We
.
assume that
for
may then
C) over the
ex ante contract between
is
could be contracted upon; the "reduced
disappear).
sold ex post at price Vi to a third party,
— cti
E
RU
resulting
and
C
37
.
and the sharing
income can be
fully specified
In the absence of subsequent
innovation, ownership rights over the joint innovation then do not matter per
se since all incentives are provided
by the sharing
nological diffusion can only be beneficial to
rule.
RU, who
Furthermore, tech-
realizes a net
expected
benefit equal to anqoVi.
Once the
possibility of future innovations
based on RU's knowledge
introduced, however, transferring this knowledge to
monopoly
rents that
C may
is
dissipate future
would have otherwise accrued to RU. In particular, sup-
pose that making the future innovation requires RU's technological knowl-
With
edge.
this
knowledge, innovating costs nothing in period 2 (but
infeasible in period 1).
The second innovation has
value V^ to a third party.
In a research joint venture (RJV), both parties have the right to
second innovation. They therefore compete
ferred
36
its
RU'b
knowledge
in period
1.
In contrast,
if
is
make
the
the research unit has trans-
under
vertical integration with
is thus similar, from the point of view
impact on the discovery of the current innovation, to a choice of effort e.
However, the cost of transferring knowledge ("e = 1") is endogeneous and depends on the
allocation of property rights on future innovations, i.e. on the dynamics of the research
of
its
decision of whether to transfer knowledge
static
venture.
37
"Information about exact payments or royalty rates is often deleted from the contracts.
However, the basic structure of compensation is generally observable. Contracts (in the
biotechnology industry) employ a rich variety of payment structures: fixed
lump sum bonus and
royalty schemes" (Pisano-Mang (1992)).
37
fee, "cost plus",
C -ownership 38
(VI), the research unit
make and commercialize
is
bound by a
and can
trailer clause
the second innovation only with the assent of the
customer. [Alternatively, and almost equivalently, one could assume that
RWs
novating in period 2 requires both
trailer clause
when the
current innovation
Not
surprisingly, the analysis
We
choose to develop the
this future
knowledge and the use of the current
Ownership of the current innovation
innovation.
is
is
then equivalent to the
not needed for the second one.
and the conclusions are the same
first
as
it
is
in-
in
both
cases.
We
notationally a bit simpler.]
take
independence of the research unit to characterize research joint
ventures.
Consider
1,
RU
only
RU has
first
an RJV.
If
no technology transfer has taken place
in period
can generate the subsequent innovation; on the other hand,
transferred
its
if
knowledge to C, both parties can generate the subse-
quent innovation and thus compete on the same product market (assuming
antitrust enforcement).
Then, under Bertrand competition,
period payoff shrinks from V2 to zero
period
only
1.
transfers
if it
its
RWs
second-
technology to
Overall, this technology transfer will be incentive compatible
C
if
in
and
if:
<*iqoVi
> V2
_12_
=
,
or equivalently:
ai1
In the integrated case
the future innovation,
has transferred
its
>
"
where only
RWs
*V,
C
a *Jv
has the right to
second-period payoff
technology in period
38
1
since
is
C
make and commercialize
again equal to zero
can dispense with
if
RU
RWs
As we shall argue at the end of this section, the following analysis and results can
be straightforwarldy extended so as to include the possibility of vertical integration under
R [/-ownership.
38
services.
period
1,
On
C
the other hand,
if
RLfs technology has not been transferred
cannot dispense with RU's
(still)
private knowledge, which in
turn guarantees a second-period payoff equal to
transfer in period
will
1
in
^
to
RU
39
A
.
technology
then be incentive compatible in the VI case
iff:
or
1
Hence the
to invest is
"
V
2q
1
terms
cost of technological diffusion in
ofC
's
reduced incentives
lower in the integrated case than in the case of a research joint
venture (reduced incentives
knowledge from
product market
RU
C
to
effect). In
addition, the diffusion of technological
does not induce (Bertrand) competition on the
RJV
in the integrated case (contrary to the
case). In other
RJV
words, technological diffusion dissipates future profits in the
not in the integrated case (rent dissipation
effect).
On
case, but
the other hand, both
product market competition and reduced incentives for the customer can be
avoided in an
RJV
knowledge from
by setting
RU
a =
t
0, at
the cost of inducing no diffusion of
to C.
RJV and
Overall, the total surplus generated by an
an integrated struc-
ture are given by, respectively:
WJUt =
<T
if
.
=
the technology transfer
(* + r {E'
{qo
is
((1
- af")Vi)))
Vi
+
+ r(E'((l-a\ I )V )))V + V3
1
1
induced, and by
W£ j" = r (£"(Vi)) V, + V
d
2
M We
= Wffiff
keep assuming alternating-offer bargaining between
between two successive offers.
39
RU
,
and
C
with no delay
if
no technology transfer
is
induced. Full vertical integration unambiguously
dominates the research joint venture since
So
far,
L7V
> a\ J
a!j
we have emphasized the advantages
of fully integrated structures
between research units and
as minimizing the cost of technological diffusion
development
units.
RJV
However, an
preserves the entrepreneurial spirit
and generates more future innovations than
the integrated structure.
that by committing a noncontractible investement c
increase the value of the second innovation by an
y. Because
it is
RU
is
.
>
in period 1,
amount A, where
Suppose
RU
can
A>c>
the only unit with the technological knowledge in period
natural to assume that only
it
1,
can make this investment. The investment
can be interpreted as a jump to the following technological generation. The
second-period payoff for the research unit will then be described as in Table
RJV
No
v2 + &
diffusion
1.
VI
v2
2
A
Diffusion
Table
Under
vertical integration,
1
RU never invests since it gets at most
of the increment in the value of the innovation
is
c
>
y. So
unaffected by the possibility of investment. In contrast,
RJV
its
and
whether
it
transfers the technology or not, since
superior product allows
A>
RU
c.
one half
RVs
payoff
invests in
an
In particular,
RU to make profit A under Bertrand competition
after the diffusion of the technology.
40
If
diffusion
is
to take place, the comparison
between
RJV
and VI now
trades off (the quality of) the future innovation against the probability of
the current one and the prevention of rent dissipation:
Wg&-Wft' = ^-c-V -[r(E-((l-ar
i
2
On
the other hand, vertical integration
by a technology
is
relative
advantage of
logical progress
l
dominated
if it is
not motivated
transfer:
W™$" - WZf"
The
)V1 ))-r(E'((l-arW ))]v
RJV
A-c>
=
0.
over VI in terms of investments in techno-
appears to be consistent with the casual observation that
large integrated firms (e.g.
AT&T
or
IBM)
are often one generation behind
technologically.
Remark
1
:
Our
analysis of the
dynamic
costs
and
benefits of vertical inte-
gration (w.r.t. short-lived research ventures) has so far been limited to the
C-ownership
case. Introducing the possibility of
^(/-ownership on
all
future
innovations does not modify the above trade-off and conclusions, as shown
in
Appendix
Remark
2
:
4.
Our
analysis so far has excluded the possibility of direct equity
participations under
RJVs.
How would
the above conclusions be modified
such equity participations were allowed? First,
of
RU's technological knowledge would be
it is
clear that the diffusion
facilitated
by having
RU
equity in C, since this would tend to soften second-period competition
diffusion occurs in period
C
seems unrealistic
in
1.
many
equity,
and
let
hold
when
However, a nonnegligible equity participation
circumstances, where the sizes of
very unequal. Conversely, suppose that
C
owns a
if
fraction (1
in
RU and C are
—
a) of RU's
us look at the outcome of the Bertrand competition between
41
l
.
RU
and
C
to C. Let
below (p
if it
which follows the diffusion of technological knowledge from
p
= —
— A)
.
If /?£/
since by doing so
p
= —
in turn
is still
if
a =
to
would obtain
at
most
—A—
<
e
(1
—
imposes a
>
its
residual payoff
v ^^
-
Then RU's second period
:
p < V%+ A, which
payoff after diffusion
equal to A, but C's payoff has increased from zero to
v~+a)-
^
^^A
(
=
V2
n °^ ner words, equity participation enforces some collusion
also facilitates diffusion by reducing
it
case of no diffusion. At the
full
RU
a)p.
be an equilibrium, one must obviously have
in case of diffusion;
over
C
has no incentive to undercut
RU:
does not undercut
For p
C
charges p, then
RU
integration
is
RU
's
payoff in
same time, the advantage of equity participation
the fact that
it
does not discourage
i?£/'s
investment
in technological progress.
7
Summary and directions for future research.
Managing innovation properly
is
one of the most important challenges faced
by developed economies. This exploratory paper argues that property rights
analysis offers a conceptual framework to understand the static
aspects involved in the organization of
few of our insights.
First, research will
an integrated structure
if
lectual inputs; in contrast,
competition
pocket.
among
activities. Let us
be more
likely to
summarize a
be conducted in
a) capital inputs are substantial relative to intel-
when
and biotechnology, research
b) the customer has
R&D
and dynamic
intellectual inputs
will often
dominate
as for software
be performed by independent
more bargaining power ex
units;
ante, say because of intense
potential research teams; c) the customer has a deep
Otherwise, research activities are more likely to be performed by
42
non- integrated research units. In that case, cofinancing by an outside investor
(venture capitalist or bank)
ond,
when
may
benefit the customer of the innovation. Sec-
there are multiple innovations and/or multiple customers, prop-
must be
erty rights
split
on the basis of comparative advantage
in creating
value. This principle gives rise to shop rights, the "hired for" doctrine,
trailer clauses.
Third, [permanently] integrated structures (or direct equity
participations) tend to facilitate transfers of technologies
to development units (customers)
tractible; in contrast, the
from research units
whenever technology transfers are noncon-
temporary nature of research joint ventures tends
more favourable environment
to create a
and
for
speeding up or increasing the
quality of future innovations, by encouraging additional specific investments
by the participants to the current research venture.
Besides these testable hypotheses
First, predicting
we obtained two "negative
results".
whether an independent research unit wants to pursue more
drastic innovations than an integrated one requires detailed knowledge of the
industry,
and
in particular of the appropriability ratio.
Schumpeterian hypotheses
fail
Second, traditional
to recognize that changes in the value of the
innovation bring about changes in the governance structure that, again, are
hard to predict without detailed knowledge of the industry.
Furthermore,
input measures often covary negatively with output measures
when property
rights are reallocated.
The
potential of the property rights analysis for studying innovation
agement
is
broader than
is
suggested by the limited scope of this paper. First,
we have considered cofinancing by a customer and
eral customers; that
is,
man-
we have not considered
investors, but not
by sev-
"horizontal" and "unrelated"
research joint ventures, where several customers competing or not in a prod-
uct market join forces to finance research. Their study will raise a host of
43
fascinating issues concerning free riding, the allocation
among customers
of
ownership rights as well as control rights over the research process, and anSecond, government promotion of
titrust policy.
R&D
one of the most
is
important areas of public policy. Analyzing the government as a customer,
an investor or a benefactor (depending on the circumstances) ought to shed
light
on
efficient
ways of channelling government money into
have not considered competition among research teams.
to
merge the property
rights
approach of
this
It
R&D.
Third,
would be
we
fruitful
paper and of the literature on
strategic vertical integration together with the traditional patent race analysis.
Last,
we
believe that our analysis provides
some microfoundations
for
extending the new growth literature in several interesting directions. In particular
it
may
help introduce financial and organizational considerations into
the "neo-Schumpeterian" framework and also enrich our current views on
technological innovation and diffusion within sectors
determinants of productivity growth.
44
and industries
as
major
References
P., and P. Bolton (1992), "An Incomplete Contract Approach to Financial Contracting", Review of Economic Studies, 59:473-494.
Aghion,
Aghion,
P.,
and
P.
Howitt (1992), "A Model of Growth through Creative
Destruction", Econometrica, 60:323-351.
Barton, J., Dellenbach, R. and P. Kuruk (1988), "Toward a Theory of Technology Licensing" in Stanford Journal of International Law, "Symposium on the Law and Economics of International Technology Licensing", 25:195-229.
Binmore, K. (1981), "The Nash Bargaining Theory", London School of Economics, Discussion Paper.
Binmore, K., Rubinstein, A. and A. Wolinsky (1986), "The Nash Bargaining Solution in Economic Modeling", Rand Journal of Economics, 17: 176188.
J. (1982), "Joint Ventures
view, 95:1523-1590.
Bradley,
and Antitrust Policy", Harvard Law Re-
J. P. Killing (1983), "The Imperfect Market for
Technology Licenses", Oxford Bulletin of Economics and Statistics,A5:223-
Caves, R., Crookell, H. and
248.
Chandler, A. (1990), Scale and Scope,
MIT
Press.
W. and R. Levin (1989), "Empirical Studies of Innovation and Market Structure" in R. Schmalensee and R. Willig, eds. Handbook of
Industrial Organization, North Holland.
Cohen,
Cohen,
W. and
D. Levinthal (1990), "Absorptive Capacity:
A New
Perspec-
on Learning and Innovation", Administrative Science Quarterly,
March,128-152.
tive
and
opment:
Fisher, F.
P.
Temin
(1973), "Returns to Scale in Research and Devel-
What Does
the Schumpeterian Hypothesis Imply?", Journal
of Political Economy,81: 56-70.
Galbraith, J.K. (1952), American Capitalism: The Concept of Countervailing Power, Boston, Houghton- Muffin.
Gale, D. and M. Hellwig (1985), "Incentive Compatible Debt Contracts:
One- Period Problem", Review of Economic Studies, 52:647-664.
The
and D. Newbery (1982), "Preemptive Patenting and the Persistence of Monopoly", American Economic Review, 72:514-526.
Gilbert, R.
45
Grossman, S. and O. Hart (1986), "The Costs and Benefits of Ownership:
A Theory of Lateral and Vertical Integration", Journal of Political
Economy,^, 691-719.
Hart, 0. and J. Moore (1991), "A Theory of Debt Based on The Inalienability of Human Capital", forthcoming, Quarterly Journal of Economics.
Henderson, R. (1991), "Underinvestment and Incompetence as Responses to
Radical Innovation: Evidence from the Photolithographic Alignment
Equipment Industry", mimeo, Sloan School, MIT.
P. Milgrom (1991), "Multitask Principal-Agent Analyses:Incentive Contracts, Asset Ownership, and Job Design", Journal of
Law, Economics and Organization, 7:24-52. (special issue).
Holmstrom, B. and
Holmstrom, B. and J. Tirole (1991), "Transfer Pricing and Organizational
Form", Journal of Law, Economics and Organization, 7:201-228.
Katz, M. and Ordover, J. (1990), "R&D Cooperation and Competition" Brookings Papers on Economic Activity, Microeconomics, 137-204.
Kogut, B. (1988), "Joint Ventures: Theoretical and Empirical Perspectives",
Strategic
Management
Journal, 319-332.
Laffont,J.J. and J. Tirole (1993),
and Regulation, MIT Press.
A Theory
of Incentives in Procurement
Nelson, R. (1959), "The Simple Economics of Basic Scientific Research",
Journal of Political Economy, 67: 297-306.
Neumeyer,
F. (1971),
policies,
Law and
The Employed Inventor
Practice,
MIT
in the
United States:
R&D
Press.
Pisano, G., Shan, W., and D. Teece (1988), "Joint Ventures and Collaboration in the Biotechnology Industry" in International Collaborative Ventures in U.S. Manufacturing, D. Mowery, ed., Ballinger Publishing Co.
Pisano, G., Russo, M., and D. Teece (1988), "Joint Ventures and Collaborative Arrangements in the Telecommunications Equipment Industry" in
D. Mowery, ed. International Collaborative Ventures in U.S. Manufacturing, Cambridge, Ballinger Publishing Co.,p23-70.
Pisano, G., and P. Mang (1992), "Collaborative Product Development and
the Market for Know-How: Strategies and Structure in the Biotechnology Industry", mimeo, Harvard Business School.
Reinganum,
J. (1983), "Uncertain Innovation and the Persistence of
poly", American Economic Review, 73:741-748.
Mono-
Rodriguez, D. (1992), "The Growth and Diffusion of Knowledge and the
Theory of the Firm", mimeo, MIT.
Rubinstein, A. (1982), "Perfect Equilibrium in a Bargaining Model", Econometrica, 50:97-110.
46
Schumpeter,
J.
(1942),
Capitalism, Socialism and Democracy,
New
York,
Harper.
Stahl,
I.
(1972), Bargaining Theory, Stockholm School of Economics.
Townsend, R. (1979), "Optimal Contracts and Competitive Markets with
Costly State Verification", Journal of Economic Theory, 21:265-293.
von Hippel, E. (1988), The Sources of Innovation, Oxford University Press.
47
Appendix
1:
The ambiguous
effect of
a change in the value
form
of innovation on the organizational
Example
(an increase in
1
V
makes
RU
ownership more desirable).
Suppose that the probability of discovery depends only on the research
unit's effort:
r(E)
that q(e)
= 2e/VQ
chooses
=
e
ownership
V
C
prefers
if
E. Then
for all
e
<
and
1,
q(e)
E m = Ec =
= 2/V
RU ownership,
ownership to pure
V>
to cofinancing. For
RU
V
jumps from
Example 2 (an increase
Assume
V
in
q(e)
V>V
,
y—e <
for all e
r
)
made
y >
RU
V>
>
0.
r V/2,
ownership to
roV. Cofinancing
as small as possible
by
We
thus conclude that at
C
RU
V
ownership.
ownership more desirable).
that the customer's and the investor's investments are not sub-
= 2e/V
<
for e
is
is
useless.
Vq/4 and q(e)
=
nonmonetary and noncon-
We
posit that r(E)
1/2 for e
>
Vq/4.
=
We
1\i\fE
assume
.
Under
C ownership,
E, yielding payoff fi^V 2
C
V< V RU
For
1.
prefers
ownership to
makes
tractible while investor's investment
and
C
Indeed, the customer's investment
stitutes.
Suppose further
(any nonnegligible amount of equity taken
by an investor then destroys RU's incentives).
the organizational form
C
+
£-
ownership, but can be
arbitrarily close to
0.
ownership since r
Vo,
f
>
for e
since |f
RU
ro is sufficiently small, since
then improves on
choosing
<
for
and more generally
C
>
r
even under pure
C
Therefore
=
chooses
E
so as to
the customer chooses
.
E so as to maximize 2f*vE V —
Under pure RU's ownership,
maximize
^
+ 2fi^/E~\ y — E,
RU
chooses e
yielding payoff
Suppose these two payoffs are equal:
+ ^y- =
2
/i
V2
<=>
48
V=
V=—
= V /4.
j+
'v*
lL
j-
Then
dV
Thus an increase
in
V
makes
C
C
V
1
VmV .
4
= - >
(and give a lump
As
in
Any
V=
V*.
example
sum payment
but cannot really be distinguished from pure
close to Vo-
0.
4
ownership optimal at
ruled out cofinancing.
RU
investor take equity in
for
2
21/2
fi
4
we have
Until now,
V
+
li'V-
RU
nonnegligible equity of an investor in
to
1,
C)
having an
optimal
is
ownership
RU
if
V
is
then destroys
i?f/'s incentives.
Last, note that at
V=
^ > pc
V", p 1
.
Therefore an increase
in
V
V*
at
discontinuously reduces the probability of discovery, while discontinuously
raising the
measured input (E).
Appendix
2: Credit constraints
novation
We
:
and the management of
an example
consider the situation where
C
has no cash to finance
its
(i.e.
>
0.
when
We look for an
Cq
=
0),
example where,
i?C/-ownership
is
in the
absence
(weakly) dominated
by C-ownership whereas the opposite holds once credit-constraints are
duced
(co
>
intro-
0).
We assume the following timing.
and specify an investment
under our assumptions).
and a bank
in-
RU and C) cannot verify the customer's profit unless they
incur a verification cost Co
of credit-constraint
E
investment
and, following Townsend (1979) and Gale-Hellwig (1985), where outside
vestors (other than
in-
E
C has
First,
C and RU allocate property rights
(cofinancing will turn out to be suboptimal
the bargaining power at this stage. Second,
B secretly write a credit contract
investment E. Third,
RU chooses its effort.
49
(see below) that provides
Fourth,
if
C
C with
the innovation occurs,
RU ownership, each of RU and C gets to make a take-it-or-leave-it offer
under
to the other party with probability 1/2.
C
ownership).
announces
its
Fifth,
C
(C
uses freely the innovation under
consumes the innovation
who then
net income to B,
audits
purchased).
(if
Sixth,
C
the credit contract specifies
if
so.
A
few words of explanation are
Townsend and Gale-Hellwig, a probability
announcement
/?i(7r,7r) if
game
and a reimbursement
Ro(tt)
possible
commitment
offer licensing fee
can
The bargaining
C
V
infinite- horizon
we assume
=
secrecy of the credit contract to avoid
on Rlfs
effects of the credit contract
when making the
R; i(0)
=
k
restrict attention to
incentive compatibility
(1
-
C ownership,
when
x Q )V
incentives).
ir
0,
=
Ro(0)
or
= V
V
=
and
without
i?i(0,7r)
auditing cost
is
(1
-
=>
p(0,
=
=
=
R/V.
co
=
is
E.
thus strictly positive.
(>
{
<Jo
=
*"
e
for
e
50
q(e)
+ r(E),
l^
>
ir
for all
loss of generality).
x
E))x
Consider the following example: p(e,E)
,(e)
kind x{V)
requires that
=V-R
E)R -
RU
offer.
the bank's participation constraint
p(0,
The expected
>
x
bargaining
when making
offers a licensing fee equal to
restrict ourselves to credit contracts of the following
and Ro(V)
(we can
no audit takes place, and
if
paper because we wish to avoid having an
the offer (note here that
Under
of audit x(tt) function of C's profit
from the alternating-move bargaining game assumed elsewere
with a finite-horizon audit.
We
credit contract specifies, as in
the audit takes place and reveals true income n.
difffers
in the
jr,
A
in order.
go,
where
=
ir
Indeed
and
'
E
r
=
r{E)
for
r
1
E<
E>
for
1
1,
where
l/V
<r <2/V
and
In the absence of borrowing constraint (cq
achieve under
C ownership
max
(since
RU
is
E V - E) =
{r
V/2
=
0),
+
<l.
r
maximum
the
payoff
C
can
:
does not exert any
incentive to invest (as r
q
1),
V-
Under pure
effort).
<
r
1
RU
C
ownership,
has no
and can obtain
V
[Note that cofinancing destroys RU's incentives and
nated by
C
C
ownership.] Suppose
ownership
in the
C
is
indifferent
is
bounded away from
ownership
is
Example
2:
r
V—
1.
unchanged because
Appendix
3:
between
absence of credit constraint: go
troducing costly state verification implies that
Two
On
C
is
thus weakly domi-
RU ownership and
y = tqV —
in-
C payoff under C ownership
the other hand, the payoff under
RU
needs no monetary investment.
other derivations of appropriability ratios
Process innovation in the Hotelling model
:
Consider the stan-
dard Hotelling model: The incumbent faces marginal cost
at the left
Then
1.
Co
and
is
located
end of a unit segment; consumers are uniformly distributed along
the segment, incur transportation cost
tical willingnesses to
pay v
for
t
per unit of distance and have iden-
one unit of the product sold by the incumbent
51
and perhaps an entrant; the potential entrants are located
We
the segment.
assume that v
the market, and that
t is
sufficiently large that
is
at the right
end of
a monopolist covers
consumers derive positive
sufficiently small that all
surplus in a duopoly situation, in the relevant range of process innovations.
As
in
example
1,
we
-
=
(v
=
denote the new marginal cost, so 7
let c
Co/c.
We
have:
ir?(c)
- -
c)
t
-
(v
-t-
2
= }- [t+ ^]
= Co — c — t
7n(c)
The
1
ttJ
for
>
t
t
outside option
t
t
>
<
is
£
j =^r
9il
£B
^
£
=> r
for
t
<
gsi
£
=
Co)
^
-
Co
c
(the firms share the market)
(the entrant corners the market).
f
always binding
>
(r
1/2).
Furthermore
is
increasing in
c
(that
is,
decreasing in
7)
is
decreasing in
c
(that
is,
increasing in
7)
In the Hotelling model, an independent research unit pursues less drastic
innovations than
may
but
its
integrated counterpart in the range of small innovations,
pursue larger innovations
infinitesimal innovations, ^i(c)
independent research unit
is
in the range of larger innovations.
infinitely larger
may want
to
make
whereas an infinitesimal innovation has
than tJ"(c)
—
7r™,
For
and the
sure the innovation takes place,
little
value socially and therefore
would not be pursued by the incumbent. In the case of large innovations
(t
<
£fl
^
£
),
or equivalently
size of the innovation,
to escape
it
Example
lysis
when the
competition
is
differentiation
small relative to the
very strong and the research unit tries
through a drastic innovation
3:
is
Product innovation and
l
.
vertical differentiation
:
A similar ana-
can be performed for a product innovation that raises quality from the
The reader may wonder whether this result does not contradict the result in the first
example
when t is infinitesimal. When t is close to 0, r = 1
is close to 1, as it is in
r
Co —
example 1 when the demand is very inelastic. In example 1, the elasticity of demand makes
r increasing in c. In example 2, demand is completely inelastic, and a small differentiation
makes r decreasing in c.
l
'
52
—
'
7 >
existing level 5 to level 75, where
have surplus 0s
—p
for
The
a taste parameter.
1.
For example, individual consumers
one unit of quality
set
up
is
purchased at price
If
the distribution of
the monopolist covers the market for both quality levels, and
There
more
is
where
if
1
is
and
2,
uniform,
is
the duopoly
show that
solution gives positive market share for both firms, one can
constant.
p,
otherwise the same as in examples
with Bertrand competition in case of entry.
if
s
then no distortion in the choice of research
line.
r
is
But
generally, r can be increasing or decreasing in vertically differentiated
markets.
Appendix
4:
Dynamic management: i?£/-ownership
Let us consider RU-ownership in the framework of section
by committing a noncontractible investment
c
>
0,
both
6.
RU
tentially increase the value of the second (future) innovation.
Suppose that
and
C
More
can po-
precisely,
let
us assume that with probability 1/2 the research unit alone can increase
V2
by the amount
increases
V2
technology1
.
by the same amount
The
with probability 1/2 the customer alone
Ac = A
Then, assuming that A/2
for second-period
2
Ajm = A, and
>
c
>
provided
does transfer
its
A/4, Table 2 gives the matrix
expected payoffs.
following analysis can easily be extended to the
investments in Vj by
RU
RU
and
C
more general case where
successful
are not mutually exclusive, provided the two successes
are not perfectly correlated.
53
RJV
TRU =
No
Vi
V/c
+y
•
=
t«7
w
v/
•
-f
diffusion
*«-* + *
=
C
7T
TKC/
A
—
— y
Tfic/
=
C
=
=
^2
C
=
7T
""/it;
+y
Diffusion
7T
-
C
f+y
7T
Table 2
Under C-ownership (V7c)
>
RV&
y =
expected benefits of investment axe
RU
will not invest in equilibrium.
^. Since j < c,
Similarly, under /?[/-ownership {VIru), C will not invest in improving
the future innovation.
equal to |
If diffusion is
•
to take place,
we thus obtain the same
trade-off as before be-
tween (the quality of) the future innovation and both the current innovation
and the prevention of rent dissipation 3 Namely, the benefit of /?[/-ownership
:
and C-ownership over
RJV
is
a lower cost of technology transfers and the
avoidance of ex post competition,
tion.
On
i.e.
rent dissipation over the future innova-
the other hand, investments in the quality of the future innovation,
respectively by
RU and
C,
may be
discouraged under C- or ^[/-ownership
*
On the other hand, RU -ownership is equivalent to RJV if Aru = Ac = A and there
no technology transfers.
4
Note that RU -ownership dominates C-ownership since technology transfers are costRV =
less in the former case (a
< a\ Ic < a^JV ). C-ownership could dominate
1
RU -ownership if quality emprovements by the customer were more important than those
made by the research unit.
3
are
'2267
0^9
54
MIT LIBRARIES
3
IDflD
QQ65b7DD
7
Date Due
o
'
_.
'&&
.-! ;
5
|
<
lis*"?
c.'
J
Lib-26-67