165

Why do firms do basic research

(with their own money) 7. *
Nathan ROSENBERG
Department of Economws. Stunford Untoerstty, Stanford. CA 94305, U S.A

Fmal version recewed May 1989

1 prepared to spend money for merely sentimental

or humanitarian purposes. Let us assume that they
The question to be addressed is: Why do private will spend their own money on basic research only
firms perform basic research with their own mo- when they are reasonably confident that it will
ney? Interest in this question derives from both yield a rate of return on this investment in the
analytical and utilitarian considerations. There is generation of knowledge that is at least compara-
empirical evidence in the United States, which ble to the rate of return that they would expect on
provides the main context for this paper, support- some other form of investment in more tangible
ing the view that basic research makes a signifi- capital.
cant contribution to the productivity growth of It is important to think of basic research by
the economy [4,7]. It is widely held that social private industry as a form of investment, and
returns from basic research are significant and more will be said about this later. But it may be
higher than private returns and it is for this reason useful to anticipate some of that later discussion
that most such activities continue to be financed here by pointing out that basic research, from a
by the taxpayer. This also implies that measures private firm’s point of view, is not only an invest-
aimed at increasing basic research by the private ment. It is, more precisely, and really by defini-
sector will be welfare improving. In the United tion, a long-term investment. There is. typically.
States, the federal government in the years since no reasonable expectation that these expenditures
the Second World War has provided the vast will begin to generate a cash flow in the next few
majority of all funds devoted to basic research. years or so.
Although the federal share has been declining in Why, then, should private industry be willing to
recent years, and although that share is at its make such expenditures? The question is a crucial
lowest level in about 20 years, it still constitutes one for the academic economist as well as for
about two-thirds of the total [lo]. policymakers in both the public and private sec-
If the goal is to encourage the private sector to tors. Since the seminal papers by Arrow and Nel-
spend more money on basic research, it is neces- son [2,11], it has been accepted by most economists
sary to start out by asking why they would want that a private enterprise economy fails to provide
to do so in the first place. Suppose it is taken as adequate incentives for investment in knowledge
axiomatic that private industry is in business solely production. There are several reasons for this
to make money, and therefore that firms are not assertion. First, there is inherently a high degree of
uninsurable risk and uncertainty that increases as
* The author has dewed much benefit from conversations we move along the basic research end of the
wth Harvey Brooks and &chard Nelson. He also wishes to research spectrum. Secondly - and this is specific
acknowledge valuable comments on an earher draft by to knowledge as a commodity - it is believed that
Ashlsh Arora, Marvm Chodorow, David Mowery. Ed
knowledge, once produced, is in some meaningful
Steinmueller and two anonymous referees.
sense “on the shelf’. As a consequence, neoclassi-
Research Pohcv 19 (1990) 165-174 cal/mainstream economics held that, once pro-
North-Holland duced, knowledge was freely available to all, in-

004%7333/90/$3.50 c1 1990, Elsewer Science Pubhshers B.V. (North-Holland)

166 N. Rosenberg / Why do forms do basic research?

eluding those firms that may have made no contri- ties to improve their efficiency is obviously subop-
bution whatever to the production of the knowl- timal.
edge: A classic “free rider” situation. In certain This is the major conclusion of economic the-
cases. one could counteract this by creating ory: Market incentives are insufficiently strong to
property rights in knowledge: but not all kinds of generate the socially optimal amount of invest-
knowledge are patentable in such a way as to ment in research - because of nonappropriabili-
preclude a competitor from exploiting that knowl- ties and uncertainties. However, attempts to al-
edge. For these reasons it has been held that there leviate that problem - by allowing firms to ap-
was a serious problem of appropriability, in that propriate the findings of research - create an
firms financing the research have no adequate equally serious problem because they impose re-
recourse or mechanism for appropriating the be- strictions upon the use of valuable knowledge that
nefits of the research to themselves. has already been produced.
It is important to note that this outcome is not The economist’s conclusion that normal market
the result of insufficient or imperfect competition. forces do not provide strong incentives for the
Although it is a fair charge against the main performance of research. especially basic research,
tradition of modern microeconomics that it tends is quite consistent with observations of the real
to attribute almost any problem in resource misal- world. The obvious empirical fact is that the over-
location to insufficient competition, this is not one whelming majority of private firms do not finance
of those cases. It has been pointed out that the the performance of any basic research. Basic re-
market for knowledge is inherently imperfect and search is, in fact, highly concentrated in two senses:
thin because, in order to determine the value of (1) the great bulk of all such research is performed
information, it would be necessary, in general, to in a very small number of industrial sectors, and
know the information. This of course creates a (2) within these sectors there is a handful of firms,
fundamental difficulty because, once the buyer typically large firms, that dominate the basic re-
knows the information, she has no incentive to search picture.
pay for it. Only limited data are available with respect to
Arrow has pointed out that society would in- company spending on basic research. With respect
vest insufficient resources in research even under to sectoral concentration, incomplete data pub-
perfectly competitive conditions [2]. He also argued lished by the National Science Foundation [lo, p.
that a monopolist’s pre-invention monopoly prof- 591 indicate that, in 1984, 61 percent of company-
its weakened the incentive to invent as compared financed basic research was in four sectors.
to the competitive situation [2. pp. 175-1791. Chemicals $ 677 million
Indeed, in one sense competition even exacerbates Electrical equipment 450
the problem. The likelihood that competitors will Aircraft & Missiles 248
quickly exploit the useful new knowledge weakens Machinery 209
even further the incentives on the part of competi- subtotal $1584/$2578 = 61 % ’
tive firms to invest in knowledge production in the
first place. The only way to strengthen those
incentives is by offering the firm that conducts the 2
research a proprietary control over the valuable
findings that the research generates. But such con- The question persists: Why do the firms that do
trol ~ e.g.. in the form of patent rights - puts us basic research do it? Alternatively, why do some
on the other horn of the efficiency dilemma. This
is because, once knowledge has been produced, it ’ The sectoral breakdown for a// company-funded R&D m
is costlessly available for other firms to utilize as 1984 was as follows:
well. Any restriction on such use is socially subop- Electrical equipment $ 9157 rmlhon
Machmery 8455
timal because it would deprive some firms of
Chermcals 7802
opportunities to raise their productivity by making Motor vehcles and eqmpment 5413
use of knowledge that is already produced and Prof. and scientlflc mstruments 4250
therefore available to society at no incremental Subtotal $35077/48065 = 73%
cost. To deprive firms of truly costless opportuni- [IO. p. 561.
 N Rosenberg / Whx do firms do hasrc reseurch? 167

firms find it profitable (or expectedly profitable) the extent that the findings of basic research can
to do it? be translated into patentable assets farther down-
Economists, as we have seen, have stressed the stream. first-movers may be able to consolidate
problems associated with appropriability as the their market position through patent protection.
main deterrent to basic research - the difficulty, Furthermore, buyer switching costs may be sig-
under normal market conditions, of appropriating nificant and may constitute a significant form of
the benefits generated by the research findings. protection against competitors for firms that are
But it should be noticed, first of all, that this first to enter the new product line.
requires an important qualification. Even if a firm’s Although first-mover advantages may thus be
basic research generated many benefits that it substantial, there may also be substantial first-
could not appropriate, the mere existence of such mover disadvantages as well as late-mover ad-
nonappropriabilities is never an adequate explana- vantages. If there are significant spillovers of
tion for the reluctance to perform basic research. knowledge between firms. then a late-mover could
So long as the performing firm can capture some gain the same knowledge at a lower cost while, at
of the benefits, that might be sufficient for it to do the same time. avoiding the major mistakes that
some research. It is not necessary to capture all of the first-mover made en route. Nevertheless, for
the benefits - indeed, it would be undesirable if it present purposes I wish merely to assert that
did. Research is socially desirable precisely be- first-mover advantages may frequently survive the
cause it often generates such widespread and in- offsetting disadvantages and serve as a significant
discriminate benefits. All that is necessary is that incentive for the performance of basic ressearch
market forces allow the firm to capture enough of [61.
these benefits to yield a high rate of return on its
own investment in basic research.
Thus, the existence of spillovers and nonap- 3
propriabilities that allow competitors a free ride is
not a decisive case against the performance of One fairly obvious, but nonetheless important
R&D (or specifically basic research) by private statement about basic research in private industry
firms. If the production of new knowledge gen- is that most firms that have engaged in it have had
erates commercial opportunities to the performer, fairly strong and well-entrenched positions of
the relevant calculation involves, not the size of market power. Precisely because the potential pay-
the spillovers. but whether the performing firm off to basic research is so long term, only firms
can capture enough of the benefits generated to that were reasonably confident of being around in
yield a high rate of return on its investment. Even the long term would be likely to consider the
in the extreme case of basic research, where there possibility of making such commitments. Thus.
is no prospect of establishing proprietary control the most successful basic research labs have been
over the research findings, commercial benefits in firms with strong market positions: Bell Labs
may nevertheless be very great. (especially before divestiture). IBM. DuPont, Dow
These potential benefits largely take the form Chemical, Eastman Kodak, etc. GM is reputed to
of what are called “first-mover advantages”. This have done quite a bit of basic research some years
is a big subject that will not be addressed here ago, but the commitment to basic research is said
beyond indicating the categories into which such to have declined along with the decline in GM’s
advantages may fall. They include a variety of market share in the automobile industry (GM has
learning experiences. Firms that move down such had the largest total R&D budget of any private
learning curves first - whether these curves per- American firm - $3.6 billion in 1986. What pro-
tain to cost reductions or performance improve- portion of that amount is basic research is not
ments - may be able subsequently to exploit the public information. but it is probably very small).
advantages conferred as a barrier to the entry of Another obvious reason why small firms hardly
new firms. First-movers may be able to acquire ever do basic research is that if research findings
valuable assets - e.g., of a geographic location or a are difficult to patent and hence the flow of
mineral deposit whose commercial worth will be payoffs cannot be capitalised, then these payoffs
favorably affected by new research findings. To must be appropriated via the means mentioned
 earlier - essentially through incorporating the patents have. in general, provided more effective
knowledge in the form of improved goods or protection to proprietary knowledge in the phar-
processes. This means, of course, that a larger maceutical industry than in most other industries.
market share will offer the prospect of a higher In this respect the innovative “output” of small
payoff. biotechnology firms is likely to be more readily
A separate but related point is that it is not size appropriable than is the case for small firms in
alone or market power that matter. Large firms other industries.
may be more willing to undertake basic research The venture capital industry seems to be treat-
when they have a diverse range of products and ing biotechnology as a kind of lottery. The vast
strong marketing and distribution networks that maJority of firms will almost certainly be losers
increase their confidence that they will eventually when the eventual “shake-out” takes place, but
be able to put the findings of basic research to some successful innovations may yield a very high
some good commercial use. In view of the high return - as Genentech hopes will be the case with
degree of uncertainty that surrounds the outcome its tissue plasminogen activator (TPA) for heart
of basic research, it is probably very important to attack victims.
a firm to have the confidence that it will know
how to exploit new knowledge that may turn up in
unexpected places. and that it will have the com- 4
plementary assets that will enable it to do so. For
an illuminating treatment of related issues see In understanding why some private firms do
Teece [14]. basic research it is necessary to recognize that
The considerable number of small biotechnol- businesses do not live in a neat, orderly world
ogy firms appears to contradict the view that where causal relationships are always clearly de-
investment in basic research requires a strong fined and where causality always works in one
market position. It is certainly true that, in the direction only. The business environment is much
drug industry proper, the most active research more interactive. full of “feedbacks” where some
programs seem to be concentrated in the large “downstream” development reacts back upon, and
firms - Merck, Johnson and Johnson. Lilly, Pfizer, alters behavior “ upstream”. Perhaps most im-
Upjohn, etc. But in the biotechnology field a good portant, it is full of unplanned, or accidental.
deal of basic research is currently being performed developments that then turn out to have an im-
by small companies such as Genentech. The really portant set of consequences of their own.
curious aspect of these small companies is that It is essential to emphasize the unexpected and
many of them have no marketable products at all, the unplanned, even if - or especially if - it
and hardly any have more than just a few. renders serious quantification impossible. In fact,
What seems to be happening here is that the the difficulties in precisely identifying and measur-
small biotechnology firms are engaged in basic ing the benefits of basic research are hard to
research that is believed to be close to the com- exaggerate. While this might seem to be just an
mercialization stage. At the same time. it is a interesting academic point about the limits of
highly speculative game that is being financed by certain methodologies, it has important de-
venture capitalists. as well as some large firms and cisionmaking consequences. The pomt has been
wealthy individuals. who are lured by the possibil- expressed succinctly: “ProJect selection methodol-
ity of a very high payoff. ogies of a formal, quantitative nature reduce the
What appears to be driving the small firms that tendency to perform basic research” [Xl.
perform basic research in biotechnology is the Part, but only a part, of the problem is that the
first-mover advantages - or at least an expectation output of basic research is never some final prod-
that first-mover advantages may be critical. Much uct to which the market place can attach a price
of the investment in this basic research took place tag. Rather, the output is some form of new
before the recent changes in the patent law that knowledge that has no clear dimensionality. The
extended patent protection to live organisms. But output is a peculiar kind of intermediate good that
this extension of the law must certainly strengthen may be used, not to produce a final good, but
the incentives to do basic research. Furthermore, (perhaps) to play some further role in the inven-
 N Rosenberg / Why do fwms do hnsrc reseurch? 169

tion of a new final good. These connections are, But it is not necessary to go back to nineteenth
however, extraordinarily difficult to trace with any century France. Those two spectacular scientific
confidence, even ex post. But even if these difficul- breakthroughs are cited simply because they were
ties could be overcome, the problems of evaluating so spectacular. We could, instead, look at Bell
the knowledge, and of providing an appropriate Labs in the twentieth century. Back at the end of
incentive system to reward the knowledge pro- the 1920s when transatlantic radiotelephone
ducers, would appear to be insuperable. service was first established, the service was poor
because there was lots of static. Bell Labs asked a
young man, Karl Jansky, to determine the source
5 of the noise so that it could be reduced or
eliminated. He was given a rotatable antenna to
Thus, it is doubtful that business decision- work with. Jansky published a paper in 1932 in
makers often sit down and ask, in an abstract which he reported three sources of noise: Local
way: Should we do basic research? How much thunderstorms, more distant thunderstorms, and a
basic research should we do? Obviously, private third source. which he identified as “a steady hiss
firms feel no obligation to advance the frontiers of static, the origin of which is not known”. It was
basic science as such. Presumably, they are always this “star noise”, as he labelled it, which marked
asking themselves how they can make the most the birth of radio astronomy [12].
profitable rate of return on their investment. Jansky’s experience (as well as the earlier ex-
In this context, my own emphasis on the unex- periences of Carnot and Pasteur) underlines one
pected and the unplanned is deliberate, because of the reasons why the attempt to distinguish
the history of basic research in American industry between basic research and applied research is
suggests that a very large part of this research has extremely difficult to carry out in practice. Funda-
been unintentional. That is to say, basic research mental breakthroughs often occur while dealing
findings of major significance have emerged as the with very applied or practical problems. Attempt-
unplanned byproduct of the attempt to solve some ing to draw that line on the basis of the motives of
very specific industrial problem. The fact is that the person performing the research - whether
the distinction between basic research and applied there is a concern with acquiring useful knowledge
research is highly artificial and arbitrary. The dis- (applied) as opposed to a purely disinterested
tinction is usually made to turn upon the motives, search for new knowledge (basic) - is, in my
or goals, of the person performing the research. opinion, a hopeless quest. Whatever the ex ante
But that is often not a very useful, or illuminating, intention in undertaking research, the kind of
distinction. If Pasteur had been asked what he knowledge actually acquired is highly unpredict-
thought he was doing back around 1870, he would able. Historically, some of the most fundamental
have replied that he was trying to solve some very scientific breakthroughs have come from people
practical problems connected with fermentation like Carnot, Pasteur and Jansky, who thought they
and putrefaction in the French wine industry. He were doing very applied research, and who would
solved those practical problems - but along the undoubtedly have said so if they had been asked
way he invented the modern science of bacteriol- at the time.
ogy. Similarly, if that other great Frenchman. Sadi But the distinction breaks down in another way
Carnot, had been asked, some fifty years earlier, as well. We have to distinguish between the mo-
what he thought he was doing, his answer would tives of the individual scientists and the motives of
have been that he was trying to improve the the firm that employs them. Many scientists in
efficiency of steam engines. ’ As a byproduct of private industry could honestly say that they are
that particular practical interest, he created the attempting to advance the frontiers of basic scien-
modern science of thermodynamics. tific knowledge, without any interest in possible
applications. At the same time, the motivation of
the research managers who decide to finance re-
’ Carnot made this utihtarian concern perfectly clear in the
title of his short but Immensely Influential book, published in
search in some basic field of science, may be
1824. Rt@xrons sur la puissance motnce du feu et sur les strongly motivated by expectations of eventually
muchrnes propres ir d&lopper cette puissance. useful findings. Thus, Bell Labs decided to sup-
 port basic research in astrophysics because of its search in industry suggests that it is likely to be
relationship to the whole field of problems and most effective when it is highly interactive with
possibilities in microwave transmission, and espe- the work, or the concerns. of applied scientists
cially the use of communication satellites for such and engineers. This is because the high technology
purposes. It turned out that, at very high frequen- industries are continually throwing up problems,
cies, rain and other atmospheric conditions be- difficulties and anomalous observations that are
came major sources of interference in transmis- most unlikely to occur outside of a high technol-
sion. This source of signal loss was a continuing ogy context. High technology industries provide a
concern in the development of satellite communi- unique vantage point for the conduct of basic
cations. It was out of such practical concerns that research, but in order for scientists to exploit the
Bell Labs decided to employ Arno Penzias and potential of the industrial environment it is neces-
Robert Wilson. Penzias and Wilson would un- sary to create opportunities and incentives for
doubtedly have been indignant if anyone had sug- interaction with other components of the in-
gested that they were doing anything other than dustrial world. Bell Labs before divestiture is
basic research. They first observed the cosmic probably the best example of a place where the
background radiation, which is now taken as con- institutional environment was most hospitable for
firmation of the “big bang” theory of the forma- basic research.
tion of the universe, while they were attempting to The emphasis on interactions and feedbacks
identify and measure the various sources of noise suggests a way of thinking about basic research
in their antenna and in the atmosphere. Although that, I believe. is potentially fruitful. That is, the
Penzias and Wilson did not know it at the time, performance of basic research may be thought of
the character of the background radiation that as a ticket of admission to an information net-
they discovered was just what had been postulated work. This network includes a variety of informa-
earlier by cosmologists favoring the “big bang” tion flows with no particular attempt to dis-
theory. Penzias and Wilson appropriately shared a tinguish or classify into basic or applied cate-
Nobel Prize for this finding. Their finding was gories. There is a high degree of interactivity, even
about as basic as basic science can get, and it is in embracing work that goes on within the realm of
no way diminished by observing that the firm that Development as well as Research.
employed them did so because they hoped to It is worth observing that the attempt to clas-
improve the quality of satellite transmission [12]. sify research into basic and applied categories is
The parallelism between the fundamental dis- particularly hard to take seriously in some areas
coveries of Jansky and Penzias and Wilson is very and disciplines, e.g., in the realms of health.
striking. medicine and agriculture. A strict application of
the most common criterion for basic research -
research that is undertaken without a concern for
6 practical applications - could easily lead to the
conclusion that the National Institutes of Health
I have deliberately examined those instances of are not deeply involved in basic research, or that
basic research emerging out of practical and ap- current university and industrial research in the
plied concerns because they provide a valuable realm of biotechnology contain no basic research,
entry into the question of how basic research gets which is absurd.
to be carried out in private industry. It is often In conducting its resource surveys the NSF
carried out unintentionally. It is, moreover, dif- defines basic research as research that has as its
ficult to understand if one insists on drawing objective “a fuller knowledge or understanding of
sharp distinctions between basic and applied re- the subject under study, rather than a practical
search on the basis of the motivations of those application thereof.” By contrast. applied research
performing the research. In fact, I would go much is research directed toward gaining “knowledge or
further: When basic research in industry is iso- understanding necessary for determining the
lated from the rest of the firm, whether organiza- means by which a recognized and specific need
tionally or geograp~cally, it is likely to become may be met” [9]. These definitions appear to mean
sterile and unproductive. The history of basic re- that, if the National Institutes of Health directed a
 N. Rosenberg / Why do fmns do basrc research? 171

major research thrust into cellular biology to pro- this respect a basic research capability is essential
vide the knowledge necessary for the development for making strategic decisions about the future
of a vaccine against AIDS, or a cure for a specific product line of the firm and the kinds of process
form of cancer, that none of the resulting research technologies that ought to be adopted. It can also
could be classified as basic. It is difficult to see be thought of, therefore, as providing some defen-
why the determination to deal with a particular sive capability - offering protection against the
disease cannot give rise to research that provides possibility of a new, competitive product intro-
“a fuller knowledge or understanding of the sub- duced from an unexpected direction.
ject under study”, even when there is a “practical In an even more general sense, a basic research
application” in mind. Here again the introduction capability is often indispensable in order to moni-
of motives, or goals, is less than useful, as the NSF tor and to evaluate research being conducted
is forced to acknowledge with respect specifically elsewhere. Most basic research in the United States
to research in private industry. Thus the NSF adds is conducted within the university community, but
to its definition of basic research the following in order to “plug in” to these research centers and
qualification: to exploit the knowledge that is generated there, a
firm must have some in-house capability. A firm is
“To take into account industrial goals, NSF
much less likely to benefit from university re-
modifies this definition for the industry sector
search unless it also performs some basic research.
to indicate that basic research advances scien-
This point is important also in identifying a
tific knowledge ‘ not having specific commercial
serious limitation in the way economists reason
objectives, although such investigations may be
about scientific knowledge and research in gen-
in fields of present or potential interest to the
eral. As I suggested earlier, such knowledge is
reporting company’ ” [9].
regarded by economists as being “on the shelf”
A further point that needs to be emphasized is and costlessly available to all comers once it has
that there are a number of activities that are been produced. But this model is seriously flawed
essential to the success of business firms in high because it frequently requires a substantial re-
technology industries that depend heavily upon a search capability to understand, interpret and to
basic research capability, even if that capability appraise knowledge that has been placed upon
does not play a direct role in solving industrial the shelf - whether basic or applied. The cost of
problems. For one thing, firms often need to do maintaining this capability is high, because it is
basic research in order to understand better how likely to require a cadre of in-house scientists who
and where to conduct research of a more applied can do these things. And, in order to maintain
nature. Indeed, that must be a major reason for such a cadre, the firm must be willing to let them
the performance of basic research in private in- perform basic research. The most effective way to
dustry. Many firms need to have a basic research remain effectively plugged in to the scientific net-
capability because that capability is essential to work is to be a participant in the research process.
making effective decisions about their applied re- These assertions require some qualification and
search activities. For another thing, a basic re- shading. Much can be accomplished in monitoring
search capability is essential for evaluating the and evaluating many kinds of research activities
outcome of much applied research and for perceiv- conducted elsewhere by in-house personnel who
ing its possible implications. are strongly motivated and who place a high value
In providing a deeper level of understanding of upon such activities. The Japanese have effectively
natural phenomena, basic research can provide demonstrated these possibilities in the last 30 years
valuable guidance to the directions in which there or so. Nevertheless, the success of this monitoring
is a high probability of payoffs to more applied capability will often be determined by the sophis-
research. In this sense, William Shockley’s educa- tication of the in-house staff in evaluating the
tion in solid state physics during the 1930s may significance of basic research findings. Moreover,
have been critical to the decision at Bell Labs to Japan’s monitoring achievements were carried out
look for a substitute for the vacuum tube in the primarily with respect to technological knowledge
realm of semiconductor materials - a search that while Japan was still in a “catch-up” mode, rather
led directly to the invention of the transistor. In than with respect to research that was at or near
 172 N. Rosenberg / Why do firmsdo hosrc rewmh?

the scientific frontiers. For an insightful treatment Thereafter it rose steadily to an (estimated) 72
of related issues see Abramovitz [l]. percent in 1986 [9, p. 2261. Although it is not
entirely clear what the impact of this growth has
been on company-funded basic research, the larger
7 role of the military tends to reduce the importance
of basic research spending within the federal
A final factor that influences the willingness of budget. This is because defense R&D expendi-
private firms to finance basic research is the role tures are very highly development-intensive com-
of the federal procurement process, particularly pared to nondefense R&LD expenditures. Weapons
military procurement. The existence of this enor- systems involve immense development costs and
mous market obviously influences the R&D deci- skew the federal R&D budget heavily in that
sions of private firms that want to improve their direction, as the following data indicate [3].
visibility and their eligibility for government mili-
tary procurement contracts. An obvious way to do 1982 Federal R&D expenditures
(T share)
this is to signal one’s capabilities by performing
R&D of the relevant sort. This practice is directly Defense Nondefense

encouraged by government sponsorship of design Basic research 3.2 33.7

and technical competitions in which potential con- Apphed research 11.0 35 3
Development 85.8 31.0
tractors participate, at least partially at their own
expense. According to one set of estimates. in
1984 about 30 percent of R&D expenditures by Thus, the growing role of military procurement
private industry was stimulated by the prospect of plus the rising share of defense in the federal
securing government procurement contracts R&D budget may be exercising major indirect
(primarily defense) [5]. There are a number of effects upon the performance of the civilian econ-
problems with these estimates. Furthermore, for omy. It is obviously essential to examine more
our present purposes, they do not disaggregate carefully the effects of large military purchases
total R&D into separate components such as basic upon the composition of all R&D activities. in-
research. Nevertheless. they suggest that a large cluding those in what are regarded as the civilian
share of private R&D may not be directed toward sector. It is also important to examine the range of
normal commercial markets where they might activities within the huge defense R&D budget.
contribute directly to productivity growth and im- For example, even though only 3.2 percent of
proved competitiveness in domestic or interna- defense R&D is classified as basic, that is 3.2
tional markets: rather, they may be shaped by the percent of an extremely large number, and it
desire to signal the capabilities of the firm as an constitutes a significant fraction of all basic re-
attractive candidate for delivering weapons sys- search that is financed by the federal government.
tems to the federal government. In addition, the How can these basic research activities be char-
reporting of R&D expenditures inevitably con- acterized? On what categories of problems are
tains a certain amount of subjective judgment, and they concentrated? What connections do they have
the numbers reported will undoubtedly be in- with company-funded R&D? Where are they
fluenced by the desire to reduce tax liabilities. The complements and where are they substitutes? What
sharp increases in reported R&D expenditures are the prospects that the output of military R&D
that have occasionally followed quickly upon the may be new technologies of value to the civilian as
introduction of new tax incentives for R&D in well as to the military sector? It seems apparent
several countries may have been more apparent that these questions do not admit of categorical or
than real. general answers, There are reasons to believe that
The growing role of defense R&D in the federal the value of spillovers from the military to the
budget suggests that these concerns are of increas- civilian sector has changed substantially over the
ing significance. During the 1980s there has been a past 40 years or so. These spillovers may vary
major increase in the relative importance of mili- considerably, depending upon the specific com-
tary R&D. In 1980 defense R&D constituted 50 position of research projects in the defense R&D
percent of all federal R & D expenditures. portfolio. It is especially important to know what
 N. Rosenberg / W’!y do fmns do basic research7 173

are the prospects for the emergence of genuinely intellectual exercise. The point is that we have
“dual use” technologies from defense R&D been considering that portion of basic research
spending. that is financed by private industry, where deci-
In the past there have been specific civilian sions are expected to be made in terms of calcula-
technological systems that have realized substan- tions of present costs and prospective benefits.
tial benefits from military R&D. At one time or Thus, anything that strengthens the prospects for
another in the past forty years, military and space euentual financial returns is likely to strengthen
R&D have made major contributions to commer- the willingness to perform basic research. All the
cial jet aircraft (including airframes, jet engines usual forces that would strengthen the willingness
and avionics), to computers, to semiconductors, to to commit financial resources to long-term pro-
communication satellites, and to nuclear power. jects become directly relevant to decisions con-
Although it is relatively easy to identify specific cerning basic research. In fact, I would suggest
technologies where military R&D has generated that the litmus test in thinking about how to
important civilian benefits, measuring the size of influence basic research decisions through govern-
these benefits is far more difficult. Even so, there ment policy is to ask: Does this action improve
are strong reasons to believe that the nature of the prospects for deriving a financial return (even-
these spillovers is changing over time and that, in tually) from any useful products that may be
many of the industries mentioned, the connection generated by the basic research? From this point
between military and commercial research pro- of view, all government macroeconomic policies
grams has grown more tenuous. As the require- that improved the economic environment for
ments of military R&D have become concerned long-term business investment would also increase
with an increasingly arcane set of needs of modern the willingness of business to spend more on basic
weapons systems, they have moved further apart research. Basic research, in order to be successful,
from the requirements of civilian markets. In some requires the making of stable. long-term commit-
military technologies it seems that the strong em- ments. Put negatively, it is likely to be discouraged
phasis on product performance and improvement, by erratic and unexpected changes in the business
and the neglect of cost considerations. has created environment. It is favored by the reduction of
a gap that has drastically reduced the possibilities uncertainties, by increasing business confidence,
of significant spillovers to the civilian economy. and by the sense of stable future prospects, includ-
For a further discussion of spillovers, see [13]. ing the confidence that government policies them-
selves will not be subject to frequent change. Not
least important, it is favored by low interest rates
8 and reductions in the cost of capital, as is inher-
ently true of all long-term investments.
Although this paper is not primarily concerned Finally, a greater confidence in the strength of
with questions of policy, it may nevertheless be one’s downstream commercialization capabilities
appropriate to make a final observation of direct should increase the willingness to perform basic
relevance to policy. That is that basic research is, research. by strengthening the prospect that the
in a very real sense, a long-term investment and firm will capture a larger share of the potential
needs to be thought about in such terms. Basic downstream benefits that may be generated by
research represents a commitment of resources to such research.
certain present uses that may eventually have a
financial payoff, but there is an unusually high
degree of uncertainty attached to this possibility. References
What is clear is that the payoff, if it comes, is very
unlikely to come in the near term. [II Moses Abramovitz, Catchmg Up, Forgmg Ahead, and
Although thinking about basic research in the Falling Behind, Journul of Economic Hutor) (June 1986).
[2] Kenneth Arrow, Econonuc Welfare and the Allocation of
same way that an economist thinks about long-
Resources for Invention. m: The Rate and Dwectmn of
term investment in tangible goods does not ex- Inuentwe Actmiy (Princeton University Press, 1962).
haust what can usefully be said about basic re- [3] Congressional Budget Ofhce, Federal Support for R&D
search, it is nevertheless an extremely valuable and Innoc~atron (Washington, D.C.. April 1984) p. 53.
174 N. Rosenberg / Why do firm do ham rewarch 7

[4] Zvt Grtliches. Productivtty, R&D and Baste Research at [lo] Nattonal Sctence Foundatton. Natronul Putterns o/Scrmce
the Ftrm Level in the 1970s. Amerrcan Economrc Recua und Technologv Resource7 1986. NSF 86-309 (Washmgton.
(March 1986). D.C., 1986) p. 38.
[5] Frank Ltchtenberg, Prwate Inoestment m R&D to SIgnal (1 l] Rtchard Nelson, The Sample Economtca of Basic Scienttfic
Ahrh!v to Perform Gorwnment Contracts. unpubhshed Research, Journal of Pohtrcal Economy (June 1959).
manuscript, Columbta Umversity. June 1986. [12] Nathan Rosenberg, Insrde the Black Bou (Cambridge
[6] Marvm Lteberman and David Montgomery, First-Mover Umversity Press. Cambrtdge. 1982) ch. 7.
Advantages, Strutegrc Management Journal (1988) 41-58 [13] Nathan Rosenberg, Ctvthan ‘Sptllovers’ from Mtlitary R&
[7] Edwin Mansfteld, Basic Research and Producttvity In- D Spendmg: The U.S. Experience smce World War II, m:
crease m Manufacturing, Ameram Economrc Review (De- Sanford Lakoff and Randy Willoughby (eds ). Strategrc
cember 1980). Defense and the Western Aihance (D.C Heath and Co.,
(81 Howard Nason, Distmcttons between Baste and Apphed Lextngton. Mass.. 1987) ch 9.
m Industrial Research. Research Munugement (May 1981) [74] Davtd Teece, Profttmg from Technologtcal Innovatton.
24 Imphcattons for Integratton, Collaboration, Ltcensmg and
[9] Nattonal Science Foundatton. Scrence Indrutors. IY85 Pubhc Policy. Research Polrqv 15 (6) (1986) 285-305
(Washmgton. D.C.. 1985) p. 221.