Appropriating the Returns from Industrial Research and Development

TL;DR: A patent confers, in theory, perfect appropriability (monopoly of the invention) for a limited time in return for a public benefit as mentioned in this paper, however, the benefits consumers derive from an innovation, however, are increased if competitors can imitate and improve on the innovation to ensure its availability on favorable terms.

Abstract: To HAVE the incentive to undertake research and development, a firm must be able to appropriate returns sufficient to make the investment worthwhile. The benefits consumers derive from an innovation, however, are increased if competitors can imitate and improve on the innovation to ensure its availability on favorable terms. Patent law seeks to resolve this tension between incentives for innovation and widespread diffusion of benefits. A patent confers, in theory, perfect appropriability (monopoly of the invention) for a limited time in return for a public

Summary

The

• To understand how appropriability differs across industries, the authors asked each respondent to report typical experiences or central tendencies within a particular industry.
• Part 2 asked about the cost and time required to imitate innovations of rivals; the authors distinguished process from product innovations, major from typical, and patented from unpatented.
• 16 Part 3 explored the links between an industry's technology and other sources of technological contribution.
• Part 4 asked some broad questions about the pace and character of technological advance.
• This paper analyzes responses to the questions in parts 1 and 2.18.

SAMPLE CONSTRUCTION

• In the manufacturing sector, these chiefly correspond to four-digit SIC industries, although some are defined as groups of four-digit or even three-digit industries.
• An additional consideration was that F. M. Scherer's technology flow matrix, which classifies patents by industry of origin and industry of use, was also constructed at this level of aggregation.
• The questions were similar to those in Mansfield, Schwartz, and Wagner, "Imitation Costs and Patents," but covered typical rather than specific innovations.
• The number of respondents did not increase in strict proportion to the level of industry R&D or sales, but the rate of response within a line of business was not significantly correlated with industry R&D spending, sales, or R&D intensity.

METHODOLOGICAL ISSUES

• Given their interest in identifying differences in the appropriability of R&D, it is reassuring that analysis of variance confirmed the presence of significant interindustry variation in the responses to most questionnaire items.
• If two firms classified as manufacturers of industrial inorganic chemicals produce different products using different technologies, they might differ markedly in their perception of the effectiveness of patents or the time required for imitation in their "industry. ".
• To eliminate this source of heterogeneity, the authors asked respondents to identify two major innovations-a process and a productwithin their industries during the past ten to fifteen years.
• The third, and probably most important, source of intraindustry heterogeneity is the inherently subjective nature of the semantic scales used in the survey.
• In his view the technologies were fundamentally different.

Patents and Other Means of Appropriation

• Table 1 shows the pattern of responses, based on a seven-point scale, to questions on the effectiveness of alternative means of capturing and protecting the competitive advantages of new or improved processes and products.
• The first two columns report the mean response for the 22.
• Secrecy, though not considered as effective as lead time and learning advantages, was still considered more effective than patents in protecting processes.
• To preserve confidentiality, the authors do not identify any industry in which there was only one response.
• To the extent that very simple mechanical inventions approximate molecules in their discreteness and easy differentiability, it is understandable that industries producing such machinery rank just after chemical industries in the perceived effectiveness of patent protection.

Conditions Affecting Appropriability

• The patterns of covariation in the responses, however, suggested that interindustry differences in conditions affecting appropriability might be summarized by a limited number of factors.
• Such clusters could prove useful in examining links between appropriability conditions and measures of R&D, innovation, and productivity growth.
• The authors investigated this possibility by reducing the data to principal components and employing a variety of factor-analytic techniques.
• The data do not reduce very 31.
• Secrecy was important in appropriating process returns, and sales and service efforts complemented lead time and learning advantages for products.

Limitations on Effectiveness of Patents

• To understand why patent protection might be weak in some industries, the authors asked respondents to rate the importance of possible limitations on patent effectiveness.
• In particular, the lack of patentability was more serious for processes than for products, and so was the disclosure of information through patent documents.
• One initiative has been to make the legal requirements for a valid patent claim less stringent.
• The authors data identified industries in which stringent requirements for patent validity or compulsory licensing were perceived as important limitations on the usefulness of patents in appropriating returns.
• Further investigation would be required to determinejust why these two sectors appear to have difficulty establishing valid claims.

Channels of Information Spillover

• To the extent that a rival can learn easily about an innovator's technology, the incentive to invest in R&D is attenuated.
• Also Table 6 summarizes the responses to questions about the effectiveness of alternative ways of learning.
• Potential licensees may learn about the opportunity to license through patent documents, or the documents may prove useful in employing new technology once it is licensed.
• For processes, a third cluster appeared to find all mechanisms of learning relatively unproductive.
• For products this group found all mechanisms moderately effective.

Cost and Time Required for Imitation

• The correlations between the effectiveness of particular learning mechanisms and the effectiveness of alternative methods of appropriation are interesting and internally consistent.
• In particular, when patent protection is effective, learning tends to take place primarily through licensing and patent disclosures.
• The effectiveness of patents is essentially uncorrelated with the effectiveness of interpersonal channels of learning and of independent R&D, and it is negatively correlated with the effectiveness of reverse engineering.

38. With three clusters the ratio of variance among clusters to variance within clusters

• Was low, but attempts to find more than three clusters were thwarted by the persistent appearance of clusters containing only one or two lines of business.
• For each category, respondents were asked to identify (within a range) the cost of duplication as a percentage of the innovator's R&D cost.
• In these instances the relative complexity of the products presumably makes reverse engineering inherently costly despite relatively weak patent protection.
• They do not seem to be linked strongly to any other mechanism of appropriability.
• Lead time and learning advantages may permit appropriation of returns even when duplication is relatively quick and inexpensive.

R&D and Innovation

• The authors summarize how data derived from their survey have been Respondents were asked to identify, on a seven-point Likert scale ranging from "very slowly" to "very rapidly," the rate at which new processes and products had been.
• The first paper focused on the Schumpeterian hypothesis that R&D intensity and innovation rates are significantly influenced by the level of industry concentration.
• In ordinary least squares and two-stage least squares specifications that included only the four-firm concentration ratio and its square as introduced in their industries since 1970.
• Of engaging in R&D, but there was no perceptible tendency for R&D intensity to increase with size within the group of R&D performers.
• Measures derived from the survey, despite their imperfections, have also been found useful for various other purposes.

Remarks on Policy

• The authors findings suggested some general principles relevant to policies that affect the incentives to engage in innovative activity.
• To illustrate how their survey results and general perspective might inform policy discussion, consider the 1987 proposal (S. 438, H.R. 557) that patent license agreements and other contracts relating to the use of intellectual property "shall not be deemed illegal per se under any of the antitrust laws.".
• Courts have often presumed that intellectual property protection is itself evidence for such power.
• To the other good reasons for rejecting such a presumption,53 the authors add that the mere existence of a patent or other legal protection says nothing about its efficacy in a competitive context.
• Thus in deciding a case, a court should inquire into the actual competitive significance of intellectual property protection in the particular market.

Comments and Discussion

• The authors' research program will have lasting value for people interested in R&D markets and markets for intellectual property, also known as Richard Gilbert.
• Clearly, the authors would like to know how market structure and capital intensity in different industries influence the degree of appropriation and affect incentives to innovate.
• The authors really want to know whether there are systematic relations between the degree of appropriation and other observable economic variables.
• Also, I suggest that in their survey work the authors include a definition of R&D.
• There is an empirical problem with surveys of the relationship between competition and R&D.

General Discussion

• Richard Levin agreed with Zvi Griliches that the appropriability variables could not discriminate effectively among more than about ten industry groupings, but he suggested that this may be a good thing, especially in light of Richard Gilbert's concern that studies such as this amount to "picoeconomics," from which no generalizations can be drawn.
• Moreover, he added, a principal conclusion, that patents do not matter very much except in the chemical industries and in semiconductors, comes through regardless of problems with questions about other mechanisms of appropriation.
• The innovation provided by one firm makes the product invented by another firm more valuable.
• So the role patents play is to define the property rights (usually through the licensing process) so that the proceeds of this cumulative process can be shared and innovation can be encouraged.
• The section provides a procedure for pursuing complaints about unfair trade practices abroad, but these complaints must be brought product by product.

Full text

RICHARD C. LEVIN
Yale University
ALVIN K.
KLEVORICK
Yale University
RICHARD R. NELSON
Columbia University
SIDNEY
G. WINTER
Yale University
Appropriating
the
Returns
from
Industrial
Research
and
Development
To HAVE
the incentive to
undertake research and
development,
a firm
must be able to appropriate returns sufficient to make the investment
worthwhile. The benefits consumers derive from an
innovation, how-
ever,
are
increased
if
competitors
can
imitate
and
improve
on the
innovation to ensure
its
availability
on favorable terms. Patent law seeks
to
resolve this tension
between
incentives
for innovation
and
widespread
diffusion of benefits. A patent confers, in theory, perfect
appropriability
(monopoly of the invention)
for a
limited
time
in return
for
a
public
We
are grateful for the support
of the National
Science
Foundation
and especially to
Rolf Piekarz of the NSF's Division of
Policy Research and Analysis.
We also wish to thank the 650
respondents
to
our
survey
and the R&D executives
who
helped us pretest it-especially Ralph
Gomory, Bruce Hannay, and Lowell Steele. Donald
DeLuca, Wendy Horowitz, and other
members of the Roper Center for Survey Research
helped manage the survey. Robert W.
Wilson, and Margaret Blair, Marc Chupka, Emily
Lawrance, Constance Helfat, Andrew
Joskow, Kathleen Rodenrys, Somi Seong, Andrea
Shepard, and Hal Van Gieson also
provided valuable assistance.
783

784
Brookings
Paper-s on
Economic
Activity,
3:1987
disclosure that
ensures, again
in
theory,
widespread diffusion
of
benefits
when
the patent
expires.
Previous
investigations of the
system suggest
that
patents do
not
always work
in
practice as
they do
in
theory.' On
the
one
hand,
appropriability is not
perfect.
Many patents can be
circumvented;
others
provide little
protection
because of
stringent legal
requirements for
proof
that
they are valid
or
that
they
are
being infringed.
On
the
other
hand,
public
disclosure
does not
always
ensure ultimate
diffusion
of an inven-
tion on
competitive terms. For
example, investments
to
establish the
brand name
of
a
patented
product
may
outlive the
patent
itself.2 And
patents may not
always
be
necessary.
Studies
of the
aircraft
and
semiconductor
industries
have shown
that
gaining
lead
time and
exploit-
ing
learning
curve advantages are the
primary
methods of
appropriating
returns. Other
studies have
emphasized
the
importance of
complemen-
tary
investments
in
marketing
and
customer service.3
Evidence
on the nature and
strength
of conditions for
appropriability
and
on the
working of
the
patent system
is,
however,
scattered
and
unsystematic.
Because
imperfect
appropriability
may
lead
to
underin-
vestment
in
new
technology,
and
because
technological
progress
is a
primary source
of economic
growth,
it
would
be useful
to have a
more
comprehensive
empirical
understanding
of
appropriability,
in
particular,
to
identify
those industries
and
technologies
in
which
patents
are
effective
in
preventing
competitive
imitation of
a new
process
or
product.
It
would also be
desirable
to
know
where
patents
can
be
profitably
licensed. Where
patents
are not
effective,
it
would
be useful to under-
stand
why they
are
not
and whether other
mechanisms
are.
1. F. M.
Scherer and
others, Patents and the
Coiporation: A
Report
on
Industrial
Technology
under
Changing Public
Policy,
2d ed.
(privately
published,
1959);
and C.
T.
Taylor and Z.
A.
Silberston, The Economic
Impact of the Patent
System:
A
Stludy
of
the
British
Exper-ience
(Cambridge
University Press,
1973).
2. See, for
example,
Meir
Statman,
"The
Effect of Patent
Expiration
on
the
Market
Position of
Drugs,"
in
Robert B.
Helms, ed.,
Drulgs and Health:
Economic Issues anid
Policy Objectives
(Washington,
D.C.: American
Enterprise
Institute,
1981), pp.
140-51.
3. The
importance of
lead time and
learning
curve
advantages
is documented in
Almarin
Phillips,
Technology and Market
Structure:
A
Study of
the
Aircraft
Indiustry
(Lexington
Books, 1971);
and John
E.
Tilton,
International
Diffuision of
Technology:
The
Case
of Semicondiuctors
(Brookings, 197
1).
For the
importance
of
marketing
and customer
service, see
Marie-Therese Flaherty, "Field
Research on the Link
between
Technological
Innovation
and Growth:
Evidence from
the
International
Semiconductor
Industry,"
working paper
84-83
(Harvard
University,
Graduate School
of Business
Administration,
no
date).

R. C.
Levin,
A. K.
Klevorick,
R.
R.
Nelson,
and S.
G.
Winter
785
This
paper
describes the
results
of an
inquiry into
appropriability
conditions
in
more
than
one hundred
manufacturing
industries.
We
discuss
how this
information
has been and
might
be used
to cast
light on
important issues
in
the
economics of innovation and
public
policy. Our
data,
derived from
a
survey
of
high-level R&D
executives,
are
informed
opinions
about
an
industry's
technological
and
economic
environment
rather
than quantitative
measures of inputs
and
outputs.
Although our use of semantic scales to
assess,
for
example,
the
effectiveness of
alternative means
of
appropriation
introduces consid-
erable
measurement
error,
more
readily
quantifiable
proxies
would
probably not serve
as well.
Remarkable
progress has been
made
toward
developing
a
methodology
to estimate the
economic value of
patents.4
But suitable data are
as yet unavailable
in
the
United
States,
and
European
data lack
sufficiently reliable detail
to support
inferences
about
interindustry
differences
in the value
of
patents.
Ourjudgment
was that
asking
knowledgeable
respondents about
the
effectiveness of
patents
and
alternative
means of
appropriation
was at least as
likely to
produce
useful
answers as
asking for
quantitative
estimates of the
economic
value
of a typical patent.
We
have
taken
considerable
care to
establish
the
robustness of our
findings
in
the
presence
of
possibly
substantial
measurement
error,
but
ultimately the value
of
the data will
depend on their
contribution to
better
empirical
understanding
of
technological
change
and
more dis-
criminating
discussion of
public
policy.
To view the
empirical
contribu-
tion
of
the
data from the
simplest
perspective,
consider
their
potential
for
improving
the
quality
of
research that uses
patent
counts to measure
innovative
activity.5
This line of
inquiry
has
shown,
among
other
results,
that
industries
vary
significantly
in the
average
number of
patents
generated
by
each
dollar of
R&D investment.6
Our
findings
on
industry
4. See,
especially, Ariel
Pakes, "Patents as
Options:
Some
Estimates
of the Value of
Holding
European Patent
Stocks,"
Econometrica,
vol. 54
(July 1986), pp.
755-84.
5.
For a
summary of the
best
of
this
work,
see Zvi
Griliches, Ariel
Pakes,
and
Bronwyn
H.
Hall,
"The Value of
Patents
as
Indicators
of
Inventive
Activity,"
working paper 2083
(Cambridge,
Mass.: National Bureau of Economic
Research,
November
1986).
For
other
perspectives on
the
usefulness of patent
data,
see the
special
issue of Research
Policy,
vol. 16
(August 1987).
6.
F.
M.
Scherer,
"The Propensity to
Patent,"
International
Journal of Industrial
Organization,
vol.
1
(March
1983), pp.
107-28;
and
John
Bound
and
others,
"Who Does
R&D and
Who Patents?"
in
Zvi Griliches,
ed., R&D,
Patents
andPProductivity
(University
of
Chicago
Press for National
Bureau
of
Economic
Research, 1984), pp.
21-54.

786 Brookings Papers
on Economic
Activity,
3:1987
differences in patent effectiveness may help explain
this variation in the
apparent productivity of R&D.
More fundamentally, large
and
persistent interindustry
differences
in
R&D investment and innovative performance have
resisted satisfactory
explanation,
in
part for lack of data
that
adequately
represent the
theoretically important concepts of appropriability
and
technological
opportunity. Promising
but
ultimately
unsatisfactory
results have been
obtained
in
exploratory
work that used crude
proxy
variables and
econometric ingenuity
to
capture
the influence of appropriability
and
opportunity
conditions.7
Our
desire
to
provide
a
stronger
basis
for this
line of inquiry was
a
prominent motive
for our survey research and
helped to shape
its
design.
Finally, gathering
better information on the
nature and
strength
of
appropriability is particularly timely
in
view of
the
prominence
of current
debates
on the
adequacy
of
laws and
institutions
to
protect intellectual
property. One impetus
for
change
has
been
the need to clarify
and
perhaps strengthen
the
system
of
property rights
at
various
new
frontiers
of
technology. Thus,
for
example,
recent
legislation
has
adapted copy-
right law to protect the rights of the creator of new
computer software,
a
new legal framework has been constructed to
protect intellectual
property embodied
in
semiconductor chip designs,
and
important court
decisions and administrative actions
have
shaped
the
development
of a
property rights system
in
biotechnology.8
Another spur
to
change has
been the
need
to resolve conflicts between
the aims of social
regulation
and the exercise of intellectual
property
rights.
For
example,
the
Drug
Price
Competition
and Patent
Term
Restoration Act of 1984
extended
patent
lives of
pharmaceuticals
to
compensate
for
regulatory requirements
that
delay
the introduction
of
new drugs.
7. Richard C.
Levin,
"Toward an
Empirical
Model
of Schumpeterian
Competition,"
working paper 43
(Yale University, School of
Organization
and
Management,
1981);
Richard
C.
Levin
and
Peter
C.
Reiss, "Tests
of a
Schumpeterian
Model of
R&D
and
Market Structure,"
in
Griliches, ed., R&D, Patents and
Productivity, pp. 175-204;
and
Ariel Pakes and
Mark Schankerman, "An
Exploration into the Determinants of
Research
Intensity," in
Griliches,
ed., R&D, Patents and
Productivity, pp. 209-32.
8. See
Computer
Software Act of
1980;
Semiconductor
Chip
Protection
Act
of
1984;
Diamond v.
Chakrabarty, 447 U.S. 305 (1980),
holding
that
plant
and animal
life is
patentable under
U.S. patent law; and
D.
J.
Quigg,
memorandum of
April 7,
1987,
explaining the
policies of the U.S. Patent and Trademark Office
concerning applications
to patent life
forms.

R.
C. Levin,
A. K.
Klevorick, R.
R.
Nelson, and S.
G. Winter
787
Intellectual
property rights
also
figure
prominently
among
policy
issues
milling under the
banner of
competitiveness.
Recent
annual
reports of
the U.S.
trade
representative
have
focused on
the
difficulties
U.S.
manufacturers encounter
in
protecting
intellectual
property
rights
in
foreign
markets.
The trade bill
passed
in
1987
by
the
House
of
Representatives contains
several
provisions
that
increase the
scope of
protection
and the
opportunities for relief
available to U.S.
manufactur-
ers
confronted
with
imports
that
infringe
these
rights.9
Proposed
antitrust
legislation,
motivated by
a
concern that courts
have
kept
inventors
from
reaping
rewards that
patent
laws are
intended to
provide,
stipulates
that
patent
license
agreements
and similar
contracts
relating
to use
of
intellectual
property
"shall
not be deemed
illegal
per
se under
any
of the
antitrust
laws."
10
To the extent
that
all this
activity
attempts
to
rectify
obvious
inade-
quacies
in
existing
institutions,
the case for
reform
appears
strong
and
straightforward.
It
is
easy
to
deplore
the blatant
copying
of
innovative
integrated
circuit
designs,
the
importation
of
"knock off"
copies
of
trademarked
or
patented
U.S.
products,
and the
piracy
of
copyrighted
written
matter and
audio and video
cassettes.
But
reforms
may yield
unintended
consequences.
In its
simplest
form,
this
concern translates
into wariness about
Trojan horses:
provisions
brought
into the
law
by
the rhetorical
tug
of
"competitiveness" and
"intellectual
property"
may
harbor
instruments of
protectionism
and
price
fixing.
Other
potential
consequences
are subtler but
no less
important.
For
example,
seemingly
uniform
adjustments
of
intellectual
property,
antitrust,
or trade
law
may
affect some
industries quite
differently
than
others.
And it
should not be
taken
for
granted
that
more
appropriability
is
better,
that better
protection
necessarily
leads to
more
innovation,
which
yields
better
economic
performance-higher standards of
living,
better
competitiveness,
and
so
on.
Better
protection
may yield
more
innovation
at
the
cost of
incrementally
increasing
resources
devoted to
producing
the innovation: the
larger prize
may
merely
encourage duplicative
private
effort to
capture
it."I
Alternatively,
better
protection
may
induce
inno-
9. See
H.R. 3, the
Omnibus Trade and
Competitiveness
Reform Act
of
1987, which is
currently
under
consideration by a
House-Senate conference
committee.
10. H.R.
557 and S.
438,
100
Cong.,
1
sess.
11. This
is the "free
access"
externality, first
emphasized
in
the context of
innovation
in
Yoram
Barzel, "Optimal
Timing
of
Innovations," Review
of
Economics and
Statistics,
vol. 50
(1968), pp. 348-55.
For a survey
of the
literature on patent
races, see
Jennifer