Journal of Economic Perspectives—Volume 5, Number 1—Winter 1991—Pages 29–41

Standing on the Shoulders of Giants:

Cumulative Research and
the Patent Law
Suzanne Scotchmer

S
ir Isaac Newton himself acknowledged, "If I have seen far, it is by
standing on the shoulders of giants." Most innovators stand on the
shoulders of giants, and never more so than in the current evolution of
high technologies, where almost all technical progress builds on a foundation
provided by earlier innovators. For example, most molecular biologists use the
basic technique for inserting genes into bacteria that was pioneered by Herbert
Boyer and Stanley Cohen in the early 1970s, and many use a technique for
causing bacteria to express human proteins that was pioneered at Genentech.
In pharmaceuticals, many drugs like insulin, antibiotics, and anti-clotting drugs
have been progressively improved as later innovators bettered previous tech-
nologies. Computer text editors are similar to one another, as are computer
spreadsheets, in large part because innovators have inspired each other. An
early example of cumulative research was Eli Whitney's cotton gin, which was
quickly modified and improved by other innovators who seriously curtailed his
profit.1
1
Eli Whitney was very generous in disclosing details of his gin to other innovators, even beyond
what was required by patent law. Other innovators patented improvements, but after much
litigation the new patents were held to infringe Whitney's underlying patent. Whitney and his
partner did not recover sufficient damage awards to compensate them for their litigation and the
time that it took to enforce the patent. For an extensive discussion, see Jeanette Mirsky and Allan
Nevins, The World of Eli Whitney (Macmillan Co, 1952). Klemperer (1990) also discusses the cotton
gin.

• Suzanne Scotchmer is Professor of Public Policy, University of California, Berkeley,

California.
30 Journal of Economic Perspectives

Most economics literature on patenting and patent races has looked at

innovations in isolation, without focusing on the externalities or spillovers that
early innovators confer on later innovators. But the cumulative nature of
research poses problems for the optimal design of patent law that are not
addressed by that perspective. The challenge is to reward early innovators fully
for the technological foundation they provide to later innovators, but to reward
later innovators adequately for their improvements and new products as well.
This paper investigates the use of patent protection and cooperative agree-
ments among firms to protect incentives for cumulative research.

The Available Tools

The breadth of patent protection is a key consideration in the incentives to

innovate. Patent applicants protect themselves against competition from deriva-
tive products by claiming broad protection. Patent law would provide no
protection at all if it did not protect against trivial changes, like color or size.
The allowable breadth of claims is determined by patent examiners and the
judiciary. If broad protection is granted, then a derivative or second generation
product will likely infringe the prior patent, so a license on the original patent
is required to market it. If patent protection is narrow, then many derivative
products and applications can be patented and marketed without infringement.2
We might be tempted to conclude that broad protection encourages firms to
find fundamental technologies but discourages them from seeking out second-
generation applications and derivative products. However, these two conclu-
sions may be inconsistent, since proper incentives to find fundamental
technologies may require that the first patent holder earn profit from the
second generation products that follow. There will be no such profit if no
second-generation products follow.
Patent protection would be an unnecessary policy tool if the government
had the same information about the costs and benefits of individual research
projects as firms have. In that case, the government could simply select the
research projects that would be socially efficient and commission research from
the lowest cost firms. However, the government will generally have less infor-
mation than firms,3 and I will therefore assume that the length and breadth of
patent protection and other aspects of the government's policy toward R&D
cannot depend on firms' private information about their expected costs.4

2
U.S. patent protection is broader than that in most other countries, particularly Japan, partly due
to the "doctrine of equivalents," which can broaden protection beyond the claims in the patent
according to similarity of function.
3
Wright (1983) discusses the private information of firms as the main justification for patent
protection, rather than using prizes or contracts as incentive instruments.
4
A policy of reimbursing costs to the successful innovator would not be adequate, since a project
that was a "good bet" at the beginning might nevertheless fail.
 Suzanne Scotchmer 31

Given that the length and breadth of patent protection cannot depend on
the expected costs of an R&D project, the only way to ensure that firms
undertake every research project that is efficient is to let the firms collect as
revenue all the social value they create. Otherwise, some projects that are
socially desirable will not be undertaken. If an innovation is a reduction in the
cost of producing a good, then the social value is the saved costs. If the
innovation is an improvement to a product, the social value is the difference in
consumers' willingnesses to pay for the improved and unimproved products.
When research firms collect all the social value as profit, households still
benefit, but in their capacity as shareholders rather than as consumers.
But there are at least two problems with allowing research firms to collect
all the social surplus as profit (or as much as possible). First, strong patent
protection leads to socially inefficient monopoly pricing. Second, firms in a
patent race may overinvest in research if the patent is worth more than the
(minimum) cost of achieving it (Loury, 1979). This problem is related to the
problem of the commons: An increase in one firm's rate of investment transfers
some probability of becoming the patentholder from other firms to itself.
Because of this transfer, all firms might overinvest.5 These points are well-
recognized in the R&D literature.
When an initial innovation facilitates later ones, as is the case with basic
research, another issue arises. Part of the first innovation's social value is the
boost it gives to later innovators, which can take at least three forms. If the
second generation could not be developed without the first, then the social
value of the first innovation includes the incremental social surplus provided by
second generation products. If the first innovation merely reduces the cost of
achieving the second innovation, then the cost reduction is part of the social
surplus provided by the first innovation. And if the first innovation accelerates
development of the second, but at the same cost, then its social value includes
the value of getting the second innovation sooner.
Because of these externalities provided to later innovators, developing the
first innovation may be efficient even if its expected cost exceeds its value as a
stand-alone product. First innovators will have correct incentives to invest only
if they receive some of the social surplus provided by second generation
products. But at the same time, enough profit must be left for the second
innovators so that they will invest if investing is efficient. This essay asks how
close patent incentives can come to accomplishing that goal.
A premise in much of what follows is that firms other than the first
innovator should participate in development of second generation products.
Since the first innovator might not have expertise in all applications, more
second generation products are likely to arise if more researchers have incen-
5
But Gandal and Scotchmer (1989) show that prior agreements among firms that would otherwise
race can overcome the incentive for overinvestment, even if their research costs are private
information. When such prior agreements are allowed, the firms will invest at the efficient rates if
and only if the private value of the patent is equal to the social value.
32 Journal of Economic Perspectives

tive to consider them. In this view, contrary to the premise of much of the
patent race literature, creativity is largely serendipitous. Not every R&D firm
sees the same opportunities for new products.
However, outside research firms can integrate with initial patentholders in
at least two ways: the firms can form cooperative ventures to research and
develop new products, and they can form licensing agreements after products
have been developed and patents have been awarded. I will call these two types
of contracts prior agreements and licenses, respectively. Prior agreements
permit firms to share the costs, as well as the proceeds, of research.6 Licenses
are negotiated after research costs are sunk and patents have been awarded.
Both types of agreements can increase profit by improving efficiency and
possibly by reducing product market competition. Although many authors have
discussed cooperation in research, they have not focussed, as I will, on how the
breadth of patent protection and cooperation among research firms work
together in protecting incentives to innovate. In this view of how incentives to
innovate are protected, a key role of patent protection is that it sets bargaining
positions for the prior agreements and licenses that will form, and therefore
determines the division of profit in these contracts.

Patent Protection and Licensing

A system of property rights that might seem natural would be to protect

the first innovator so broadly that licensing is required from all second genera-
tion innovators who use the initial technology, whether in research or in
production. But such broad protection can lead to deficient incentives to
develop second generation products. When the licensing agreement is negoti-
ated after a patent has been granted, research costs have already been sunk.
The bargaining surplus to be split between the first and second innovators at
that time is the incremental market value of the second product, not net of
research costs. A second innovator who cannot market the next generation
product without a license has a very weak bargaining position. If the second
innovator does not get all the surplus being bargained over, he will earn only a
fraction of the new product's market value and presumably only a fraction of its
social value, and this fraction may be less than the cost of developing it. Hence
the incentive for an outside firm to develop second generation products can be
too weak.7 Under such broad patent protection, the incentive for the first
innovator to develop a second generation product will be stronger than for an
6
Prior cooperation in research has been treated leniently by antitrust law and the authorities. For
example, the National Cooperative Research Act (1984) established that joint ventures are not
per se illegal, but will be treated according to a "rule of reason." The Act also reduced damages in
civil suits from treble to single, provided the firms follow the proper notification procedure.
Perspectives on this act are discussed in the symposium on "Collaboration, Innovation and
Antitrust" in the Summer 1990 issue of this journal.
7
See Green and Scotchmer (1990) for an elaboration of this idea.
 Standing on the Shoulders of Giants 33

outside firm (provided the first innovator has expertise to develop the new
product, and thinks of it), since the first innovator will earn the entire incre-
mental profit.
As well as offering deficient incentives for second innovators, broad patent
protection might inefficiently inflate incentives for the first innovation. In
licensing agreements, the first innovator will earn a share of the market value
of each infringing later product. If the first innovation reduces the cost of
achieving later innovations, but is not the only possible vehicle to achieve them,
the first innovator's share should not exceed the cost reduction. If it does, the
first innovator will be overrewarded.
In what follows I explore two solutions to these defective incentives. The
remainder of this section investigates what happens if the first innovator's
patent protection is narrowed so that a different enough second generation
product does not infringe and thus can be marketed without a license from the
first innovator. In the following section, I investigate prior agreements in which
second innovators can "sell" their ideas to the first innovator or integrate with
the first innovator. Neither solution is perfect, as we shall see.
The inadequacies of narrowing patent protection are most easily exposed if
we first suppose that first and second generation products do not compete in
the market, although second generation products build on the first generation
technology; for example, many new pharmaceuticals that are therapies for
different illnesses all build on a few basic techniques of bio-engineering. Second
innovators cannot have excessive incentive to invest, since they cannot earn
more than consumers' willingness to pay in the markets they serve. Licensing
from the first innovator would transfer away some of the second innovators'
revenue and hence reduce their incentive to invest. To provide efficient
incentives to the second innovator, society should protect the first innovation so
narrowly that a new product never infringes and therefore second innovators
never have to license. But such a scheme does not sufficiently reward the first
innovator, since the first innovator does not profit from the cost reduction
conferred on the second innovators.
The first innovator's incentive to invest becomes still weaker under narrow
patent protection if the second generation product is a substitute for the first.
Competition between the two patentholders would erode their joint profit,
transferring some of the social surplus of the combined innovations to con-
sumers. As an example, suppose that the second generation product is a
superior version of a drug, and that the two patent-holders compete on price.
Then the second generation product will survive in the market and its price
will equal the difference in consumers' willingness to pay for the two drugs plus
the marginal cost of producing the drug. In this outcome, the second innovator
earns as profit exactly the incremental social value of the newer drug, while the
first innovator's profit falls to zero.
Such profit erosion could be mitigated if the antitrust authorities permitted
collusive licensing among patentholders who would otherwise compete. For
34 Journal of Economic Perspectives

example, licensing with per-unit royalties can lead to collusive outcomes, since
the royalty raises the licensees' private production cost and therefore keeps the
equilibrium price high.8 In ordinary antitrust law, collusion through licensing
would violate the spirit of the Sherman Act and subsequent legislation. But
where incentives to innovate are at stake and where later technology builds on
an earlier technology, such collusion allows the first innovator to profit from
the externality conferred on later innovators. Of course, firms would be
tempted to exploit any leniency by the antitrust authorities in contexts where
incentives to innovate are not at stake. This problem should not be minimized.
There is something quite general economists can say about the combined
effects of patent law with licensing: No such policy can achieve fully efficient
incentives, even if society permits collusive licensing between patent holders
who would otherwise compete and the firms jointly collect all the social surplus
as profit. This is essentially because of "double marginalization." To give the
second innovator an incentive to invest whenever social benefits exceed R&D
costs, the second innovator must earn the entire social surplus of his innova-
tion. But to compensate the first innovator for the externality or spillover she
provides, she too must earn part of this surplus. It is impossible to give the
surplus to both parties.9
When both first and second generation products are developed, the divi-
sion of profit between the two innovators depends on the breadth of patent
protection. To see this, assume that there is a random component to the
outcome of a research project, so that when a research firm invests in a second
generation product, it does not know whether its product will infringe the prior
patent. The breadth of the prior patent determines the probability that the
second generation product will infringe. If the second product turns out to
infringe, the second innovator must license and this will force him to share the
profit of the improvement with the first innovator. The second innovator is in a
better position if its product turns out not to infringe, since the second
innovator can profitably compete with the prior patentholder in the market.
Thus, if breadth of the first patent could be interpreted to depend on the
expected costs and benefits of a second generation product, we could ensure
that the second innovator's expected profit would be zero. If not, some second
generation products will be stymied even though they would contribute posi-
tively to joint profit and to social welfare, and the second innovators who invest
will typically make positive profit.
To summarize, the "natural" system of property rights—requiring every
later innovator to license any underlying technology—will on average give
deficient incentives for outside firms to develop second generation products.
8
If royalties are permitted, then licensing is similar to permitting the initial patent-holder to buy up
the patents on later products that use the initial technology.
9
Green and Scotchmer (1990) argue this in a model where the second innovation would be
impossible without the first innovation. It is also true in the less drastic case when the first
innovation merely reduces the cost of achieving the second innovation.
 Suzanne Scotchmer 35

This is because the second product infringes and therefore the second innova-
tor must transfer some of the innovation's revenue to the first innovator by
licensing. If the first innovator can be relied upon to develop all second
generation products, this would not matter. Second, no system of narrower
patent protection and licensing can give the right incentives to both the first
innovator and other firms that develop improvements, even if collusive licens-
ing among noninfringing products were allowed. The latter result depends on
my premise that the breadth of an underlying patent cannot be separately
tailored to the costs and benefits of each second generation product.
In the next section, I ask to what extent these inadequacies of patent
protection and licensing can be overcome with prior agreements reached
before some or all the patents have been obtained. Incentives with licensing are
defective mainly because firms negotiate after all costs have been sunk and
patents have been issued. A prior agreement integrates the potential second
innovator into the firm of the first innovator before investing in the second
innovation. Such prior agreements can indeed guarantee efficient investment
in second generation products, but cannot perfectly solve the incentive problem
unless the negotiation is before all costs are sunk, including the costs of the first
innovator, or unless the first innovator has all the bargaining power.

Prior Agreements

Prior agreements among research firms are often called research joint
ventures. Joint ventures presumably form to increase the joint profit of the
members, but they do not necessarily increase social welfare, since the coopera-
tion is among firms only and does not include consumers.10 Joint ventures
increase profit both by providing incentives to the members to invest more
efficiently, and by finding ways to transfer social surplus from consumers to
firms. The greater efficiency might result from exploiting economies of scale
(Katz, 1986), from sharing technological know-how (Bhattacharya, Glazer and
Sappington, 1988), or from undoing the inefficiencies of a patent race (Gandal
and Scotchmer, 1989).
One solution to the incentive problem would be to integrate all possible
innovators into one firm before even the first innovator has invested. Then,
provided the integrated firm gets most of the social surplus from the joint
innovations, it should invest (close to) efficiently. Although research firms do
not know with certainty what projects they will think of after the first genera-
tion technology has been developed, they have expectations about the possible
benefits and costs of such projects. Provided all researchers have similar
expectations, an agreement negotiated before the first investment could ensure
10
The Coase theorem would conclude that bargains increase the joint welfare of all the parties.
Thus, if consumers and firms could jointly cooperate, prior agreements would inevitably benefit
both groups. But since some of the parties are excluded—namely consumers—there is no
guarantee that prior agreements increase total welfare.
36 Journal of Economic Perspectives

that the first innovation is undertaken if and only if efficient, where efficiency is
defined relative to the prior judgments about costs and benefits. But the more
serendipitous is the discovery of second generation products, and the more
difficult it is to include all potential second innovators, the less feasible such an
agreement seems. I therefore consider the more limited prospects when inte-
gration occurs after the first innovation. The difficulties in transferring profit to
the first innovator are clearest if we assume that the innovators can jointly
collect all the social surplus as profit, provided they do not compete in the
market.
After the first patent has issued, a potential second innovator could ap-
proach the first patent-holder with an idea for an improvement or new
product, and suggest that they share both the costs and proceeds of research.
Such an agreement can increase joint profit by increasing investment in
profitable second generation products and by preventing market competition
among firms that would otherwise own competing patents. If patent protection
is broad, without this prior agreement the second innovator could have defi-
cient incentive to invest, as explained above. With a prior agreement, the initial
patent-holder can agree to share both the costs and the proceeds of the second
innovation, and will do so whenever benefits exceed costs.
Prior agreements are a social improvement over licensing because they can
improve incentives to invest in second generation products, whatever the
breadth of patent protection. With licensing, the breadth of patent protection
serves two purposes: It determines investment in second generation products
and determines how the firms' joint expected profits will be divided. With prior
agreements, the breadth of patent protection serves one purpose instead of
two: The two innovators have an incentive to invest efficiently in second
generation products whatever the breadth of patent protection. The breadth of
protection determines only the bargaining positions, hence the division of
profit.11
Whether a prior agreement can provide efficient incentives for the first
investment as well as the second depends on two factors: how much social
surplus must be transferred from the second innovator to the first (how big the
externality is), and the second innovator's bargaining power. A second innova-
tor who has a strong bargaining position will earn positive profit in a prior
agreement, thus limiting how much social surplus the first innovator can
collect. The second innovator's bargaining position is strongest if there is a high
probability the second innovation would not infringe and if the second genera-
tion product is itself patentable. The second innovator will also have a strong
bargaining position if no other firm is capable of developing the second
generation product.

11
On the other hand, approaching the first innovator with the idea for the second innovation might
give away the idea of the potential second innovator, and thereby undermine its bargaining
position. The law has remedies for this problem, but presumably they do not work perfectly.
 Standing on the Shoulders of Giants 37

To bid the second innovators' profits to zero, it is not necessarily enough

that each potential second innovator has many competitors. Suppose the
second generation product is itself patentable and the first innovator agrees
with one potential second innovator that that firm will develop the improved
product and have an exclusive license on the initial patent. Despite this
agreement, the parties will be forced to negotiate with any other firm that gets
the improvement first since they could not duplicate the second invention
without infringement. Anticipating this, potential second innovators might not
bid away their profit in trying to make an agreement with the first innovator
(Scotchmer, 1990).
Some authors who have written on research joint ventures have pointed
out that firms' incentives to cooperate at the research stage depend partly on
whether the members can also collude in using the resulting patents (Ordover
and Baumol, 1985; Katz, 1986; d'Aspremont and Jacquemin, 1988; Choi,
1989). My discussion has assumed they can, and that such collusion is an
important way to protect incentives to innovate. It would be difficult to
implement a rule that permitted the consolidation of property rights through a
prior agreement only if the eventual patents would otherwise be infringing.
When research outcomes are random, patent authorities cannot know whether
an R&D project will result in new technology that infringes an earlier patent.

Conclusion

A main conclusion of this essay is that it is misleading to ask how broad

patent protection should be without simultaneously asking whether research
firms can integrate or otherwise cooperate. Similarly, it is misleading to ask
how leniently society should treat cooperation among research firms without
simultaneously asking how broad patent protection should be. Prior agree-
ments have the advantage of leading to more efficient investment in second-
generation products. But if the patent authorities disallow collusive licensing of
patents that are noninfringing, they might also want to disallow prior agree-
ments. A prior agreement merges the interests of the two firms whether or not
the patents would be noninfringing, and could circumvent the antitrust author-
ity's desire to force competition. Of course, competition has a disadvantage
where incentives to innovate are at stake: It erodes the joint profit of the firms,
which undermines at least one firm's incentive to invest in R&D when such
investment is efficient.
There are no simple conclusions to draw about the optimal breadth of
patents. It is not necessarily optimal to protect the first innovation so broadly
that every derivative or second generation product infringes. If prior agree-
ments are disallowed or ineffective for some reason, then broad patent pro-
tection could discourage the development of second generation products, as
38 Journal of Economic Perspectives

explained above. And if the first innovator does not expect to profit by licensing
to second generation innovators, broad protection could inhibit the first inno-
vation as well, thus undermining the entire research line.
Broad protection might also be undesirable when prior agreements are
allowed. To encourage researchers to invest in second generation products, the
first innovator might have to make prior agreements with firms that have
bargaining power. Their bargaining power derives from the fact that, without
an agreement, they might have a credible threat not to invest, and from
patentability of the second generation product. Because of their bargaining
power, they may get a share of the bargaining surplus. Suppose patent protec-
tion is narrowed enough so that second innovators will invest without a prior
agreement, but their profit is kept low. This would increase the first patent-
holder's profit, as may be necessary to compensate the first innovator for the
externality or spillover conferred on second innovators.12
Patent law is limited in its instruments: the main ones are the patent life
and the breadth of protection.13 The private value of patent protection is
linked to the social value of the technology through market demand, but is not
linked to firms' research costs. The optimal rule for the breadth of a patent can
only use information that is available to patent examiners and courts.14 Thus,
the patenting rule can depend on observable aspects of discovered technolo-
gies, but not on prior expectations regarding technological outcomes and costs
of research. This restriction greatly reduces the effectiveness of patent law in
protecting incentives.
Before investing in a second generation technology, the researcher must
evaluate the probability that the new technology will not infringe the prior
patent. This probability depends on the breadth of the prior patent and on the
distribution of possible outcomes of the second investment. The probability of
infringing the first patent is lower if the distribution of outcomes places greater
weight on outcomes that lie outside the allowed claims of the first patent. A
project with low probability of infringing will look more profitable to the
second generation researcher than a project that places greater prior weight on
outcomes that infringe the first patent. But both projects could have the same
expected social value if the expected costs of one project were sufficiently
higher than the other. The patent policy should equally encourage two projects

12
See Green and Scotchmer (1990) for an elaboration of this argument.
13
Of course, details like priority rules also matter. Everywhere except the United States, a disputed
patent issues to the first applicant. In the United States, a disputed patent issues to the first
inventor, regardless of when application occurs. See Scotchmer and Green (1990) for a discussion
of the incentive effects of these two rules. The anomalous American rule is now being reconsidered.
14
Except for Wright (1983), authors have not focussed explicitly on what information is available to
patent authorities. Gandal and Scotchmer (1989), Green and Scotchmer (1990), Klemperer (1990)
and Scotchmer and Green (1990) assume that patent protection cannot depend on costs.
Klemperer (1990) and Gilbert and Shapiro (1990) assume that the breadth of patent protection can
depend on aspects of market demand. Green and Scotchmer (1990) assume that the breadth of
patent protection can depend on the level of previous technical advance.
 Suzanne Scotchmer 39

with the same social value, but that cannot be accomplished with a patent rule
that depends only on technological outcomes.
A disadvantage of narrow patent protection that I have not yet discussed is
that it might discourage first innovators from patenting and disclosing their
technologies. Patent law requires disclosure for the same reason that innovators
dislike it: it is the vehicle by which technical knowledge is passed from the
patenting firm to its competitors.15 The first innovator would rather develop
second generation products than let other firms develop them, since that would
be more profitable. As a consequence, the first innovator has a strategic
incentive not to patent the innovation. Instead, the first innovator could hold
the product off the market until it develops the more valuable second genera-
tion products, or it could market the first product and rely for protection on
the law of trade secrets, which does not require disclosure, but also does not
protect against independent invention.
The first innovator's incentive to patent the initial technology depends on:
(i) the profitability of marketing the first technology prior to the development
of second generation products; (ii) the extent of disclosure that patenting
entails;16 (iii) the ease with which the technology could be reverse-engineered if
marketed but not patented; and (iv) the breadth of patent protection. The
incentive not to patent is especially strong when patent protection is narrow,
since a second generation product is then more likely to damage the first
innovator's profit.
The problem of cumulative research is especially acute when the first
technology has very little value on its own, but is a foundation for valuable
second generation technologies. Even with licensing, the first innovator might
not capture the full social value that it facilitates and may have deficient
incentive to invest. This is presumably why governments fund basic research.
The branches of government that fund research are not those that set patent
policy, and the decision to support basic research might be interpreted as a
recognition that patents and licensing are inadequate.
Governments have taken different views of whether publicly sponsored
research should also be patentable. The U.S. government permits and even
encourages patenting of results from government sponsored research; for
example, the Boyer-Cohen patent. In contrast, the British government forbade
the Cambridge Molecular Biology Lab from patenting monoclonal antibodies

15
The disclosure requirement in section 112 of the patent law states that "the specification shall
contain a written description ... in such full, clear, concise and exact terms as to enable any person
skilled in the art ... to make and use the same . . . " Scotchmer and Green (1990) show that a
leading firm might not want to patent a patentable technology even if this means holding it off the
market until the next innovation in order to avoid reverse engineering.
16
Disclosure of some technologies, such as chemical compositions, teaches competitors much that is
valuable. For example, after the basic material of superconductors was disclosed, many other
researchers developed aspects of it (The New York Times, January 2, 1989, p. 34). Disclosure of other
technologies, like bio-engineered proteins, gives away much less that is useful to competitors, and
therefore innovators will not fear disclosure as much.
40 Journal of Economic Perspectives

in the mid-1970s. Permitting patents on government sponsored research re-

wards successful innovators twice, once through government funding and again
through the patents.
I have focussed in this essay on how to divide joint profit among innovators
when one innovator's technology builds on another's. I have not focussed on
the length of patent protection as a policy tool. My simplifying assumption that
the innovators can jointly get all the social surplus as profit if they do not
compete in the market essentially means that the patent continues forever.
When owners of noninfringing patents cannot collude through licensing, a
noninfringing second generation product will undermine the profitability of
the first patent, and this may happen before the end of the patent life.
Mansfield (1984) surveyed research firms and found that the effective lives of
most patents are much shorter than the stipulated 17 years. In the extreme
case where a second product is an improved version of the first, the effective
length of the first patent may simply end when the second product is intro-
duced. The effective patent life is determined directly by its breadth, since the
breadth determines how long it takes until this happens. On the other hand,
when second generation products serve different markets, as when consumers
have different tastes and a second generation product is a variant that serves a
different set of consumers, length and breadth can be chosen independently,
although the monopoly power conferred by a patent may be eroded over the
patent's life as similar but noninfringing products are invented.
Gilbert and Shapiro (1990) and Klemperer (1990) have discussed breadth
of patent protection in the context of single innovations without focussing on
the problem of dividing profit among sequential innovators. In their conceptu-
alization, breadth and length are substitute ways to provide a fixed profit to a
single innovator, and they ask what combination of length and breadth mini-
mizes the cost of monopoly distortions. They do not ask how large the fixed
profit should be and therefore do not focus on how to preserve incentives to
innovate. The prescription for patent breadth that comes out of that perspec-
tive may well conflict with the prescription for patent breadth that comes from
considering incentives to innovate when research is cumulative.
It appears that patent policy is a very blunt instrument trying to solve a
very delicate problem. Its bluntness derives largely from the narrowness of
what patent breadth can depend on, namely the realized values of the technolo-
gies. As a consequence, the prospects for fine-tuning the patent system seem
limited, which may be an argument for more public sponsorship of basic
research.

• Prepared for the Symposium on Intellectual Property Law, funded by the RAND
Corporation and the John P. Olin Foundation, which took place in Washington D.C. on
October 24, 1989. I thank Stan Besen, Joe Farrell, Jerry Green, Leo Raskind, Steve
Salop, Carl Shapiro, Eugene Smolensky, Joe Stiglitz and Timothy Taylor for useful
comments. I thank Lucette Decorde for her able research assistance and the NSF, Grant
SES 89 09503, for financial support.
 Standing on the Shoulders of Giants 41

References

d'Aspremont, C., and A. Jacquemin, "Co- Klemperer, Paul, "How Broad should the
operative and Noncooperative R & D in Duop- Scope of Patent Protection Be?" The RAND
oly with Spillovers," American Economic Review, Journal of Economics, Spring 1990, 21,
1988, 78, 1133–1137. 113–130.
Bhattacharya, S., J. Glazer, and D. Sapping- Loury, Glenn C., "Market Structure and
ton, "Motivating Exchange of Knowledge in Innovation," Quarterly Journal of Economics,
R & D Ventures: First-Best Implementation," 1979, XCIII, 395–410.
Bell Communications Research, Technical
Memorandum, 1988. Mansfield, Edwin, "R&D and Innovation:
Choi, Jay P., "An Analysis of Cooperative Some Empirical Findings." In Griliches, Zvi,
R & D," mimeo, Dept. of Economics, Harvard ed., R & D, Patents and Productivity. Chicago:
University, 1989. University of Chicago Press for the National
Gandal, N., and S. Scotchmer, "Coordinat- Bureau of Economic Research, 1984.
ing Research Through Research Joint Ven- Ordover, Janusz, and William J. Baumol,
tures," GSPP Working Paper #171, University "Antitrust for High-Technology Industries:
of California, Berkeley, 1989. Assessing Research Joint ventures and Merg-
Gilbert, Richard and Carl Shapiro, "Opti- ers," Journal of Law and Economics 1985,
mal Patent Length and Breadth," The RAND XXVIII, 331–331.
Journal of Economics, 1990, 21, 106–112. Scotchmer, S., "Protecting Early Innova-
Green, J., and S. Scotchmer, "Antitrust Pol- tors: Should Accessory Products, Bundled Im-
icy, the Breadth of Patent Protection and the provements and Applications be Patentable?,"
Incentive to Develop New Products," GSPP GSPP Working Paper #183, University of Cal-
Working Paper #171, (revised), University of ifornia, Berkeley, 1990.
California, Berkeley, 1990.
Jacquemin, A., "Cooperative Agreements in Scotchmer, S., and J. Green, "Novelty and
R & D and European Antitrust Policy," Euro- Disclosure in Patent Law," The RAND Journal
pean Economic Review, 1988, 32, 551–560. of Economics, 1990, 21, 131–146.
Katz, M., "An Analysis of Cooperative Re- Wright, Brian, "The Economics of Inven-
search and Development," The RAND Journal tion Incentives: Patents, Prizes and Research
of Economics, 1986, 17, 527–543. Contracts," American Economic Review, 1983,
73, 691–707.
This article has been cited by:

1. Duncan S. Gilchrist. 2016. Patents as a Spur to Subsequent Innovation? Evidence from

Pharmaceuticals. American Economic Journal: Applied Economics 8:4, 189-221. [Abstract] [View PDF
article] [PDF with links]
2. Fiona Murray, Philippe Aghion, Mathias Dewatripont, Julian Kolev, Scott Stern. 2016. Of Mice and
Academics: Examining the Effect of Openness on Innovation. American Economic Journal: Economic
Policy 8:1, 212-252. [Abstract] [View PDF article] [PDF with links]
3. Eric Budish, Benjamin N. Roin, Heidi Williams. 2015. Do Firms Underinvest in Long-Term
Research? Evidence from Cancer Clinical Trials. American Economic Review 105:7, 2044-2085.
[Abstract] [View PDF article] [PDF with links]
4. Petra Moser. 2013. Patents and Innovation: Evidence from Economic History. Journal of Economic
Perspectives 27:1, 23-44. [Abstract] [View PDF article] [PDF with links]
5. Jeffrey L. Furman,, Scott Stern. 2011. Climbing atop the Shoulders of Giants: The Impact of
Institutions on Cumulative Research. American Economic Review 101:5, 1933-1963. [Abstract] [View
PDF article] [PDF with links]
6. Michele Boldrin,, David K. Levine. 2009. A Model of Discovery. American Economic Review 99:2,
337-342. [Citation] [View PDF article] [PDF with links]
7. Petra Moser. 2005. How Do Patent Laws Influence Innovation? Evidence from Nineteenth-Century
World's Fairs. American Economic Review 95:4, 1214-1236. [Abstract] [View PDF article] [PDF with
links]
8. Hal R. Varian. 2005. Copying and Copyright. Journal of Economic Perspectives 19:2, 121-138. [Citation]
[View PDF article] [PDF with links]
9. Richard A. Posner. 2005. Intellectual Property: The Law and Economics Approach. Journal of
Economic Perspectives 19:2, 57-73. [Citation] [View PDF article] [PDF with links]
10. Petra Moser, Tom Nicholas. 2004. Was Electricity a General Purpose Technology? Evidence from
Historical Patent Citations. American Economic Review 94:2, 388-394. [Citation] [View PDF article]
[PDF with links]
11. Nancy T. Gallini. 2002. The Economics of Patents: Lessons from Recent U.S. Patent Reform. Journal
of Economic Perspectives 16:2, 131-154. [Abstract] [View PDF article] [PDF with links]