Public Research, Patents and Implications for Industrial R&D in the Drug, Biotechnology, Semiconductor and Computer Industries
Wesley M.Cohen
Carnegie Mellon University and John Walsh
University of Illinois at Chicago
INTRODUCTION
The links between industry and academia have deepened over the past two decades. For example, of the 1,056 estimated university-industry R&D centers existing as of 1990, almost 60 percent were established in the prior decade alone. Academic patenting activity has also increased. In 1974, 177 patents were awarded to the top 100 research performing universities. By 1995, there were 1,561. In 1980, 25 American universities had offices administering technology transfer and licensing. By 1990, the number had grown to 200. The share of academic R&D supported by industry has also increased from 2.6 percent in 1970 to 6.9 percent in 1990. Although no systematic data exist on either spin-offs or faculty participation in new firms, anecdotal evidence suggests an increase over the past twenty years, especially in biotechnology and software.1 In addition to such closer ties between industry and academia, Narin et al. have claimed that there has been a parallel increase in the commercial impact of university research, at least in selected domains.2
These deepening ties and impacts likely emerge from several sources. Academic research in a number of areas, such as in the biomedical sciences, has moved in directions that offer more immediate commercial potential. Such move-
ments are surely associated with policy decisions to provide key financial support. There are also numerous other policies and features of the policy environment that have pushed universities and industry closer together. For example, stimulated by the broad expectation dating from the late 1970s that federal support for academic research would fall, universities and particularly their faculty began to seek out more industry support for their work. At the same time, there was a shift in government attitudes toward collaborations between industry and universities. Prompted largely by growing international competition, legislative changes have encouraged academics to solicit support from industry and have also given industry an incentive to be more forthcoming. Specifically, the Economic Recovery Tax Act of 1981 extended industrial R&D tax breaks to support research at universities. In addition, since the 1970s there has been substantial growth in government programs (such as the NSF’s Science and Technology Centers and Engineering Research Centers) that tie government support for university research to industry participation.
Policy has also changed regarding the ability of universities to profit from their research and allow others to profit from it as well. The Patent and Trademark Act of 1980, known as Bayh-Dole, permits universities and other nonprofit institutions to obtain patent rights to the output of federally sponsored research. Part of the impetus behind Bayh-Dole and related legislation was the assumption that there was some stock of underexploited, commercially applicable knowledge residing in universities and other research institutions receiving federal funding. The problem, as seen by policy makers at the time, was a lack of incentive to transfer this knowledge to the private sector and subsequently embody it in product and process innovation. So, the idea was to use the award of patents to “incentivize” the private sector to undertake the downstream R&D and related investments necessary for commercialization.3 Although perhaps less critical due to the already existing incentives for faculty to publish, some suggested that patents could also benefit the public by similarly providing universities with an economic incentive to increase their technology transfer efforts.4
The conventional wisdom is that Bayh-Dole has indeed induced an outflow of commercially fruitful technology as reflected in recent increases in university patenting and licensing. Mowery et al. have, however, voiced skepticism. Reflecting on their analysis of the patenting and licensing behavior of the University of California, and Columbia and Stanford Universities, Mowery et al. conclude: “An array of developments in academic research, industry and policy… combined to increase U.S. universities’ activities in technology licensing, and Bayh-Dole, while important, was not determinative…. Even in the absence of
Bayh-Dole, we believe that all of these three universities would have expanded their patenting and licensing of faculty inventions during the 1980s.”5
At this juncture, it is difficult to test directly whether Bayh-Dole has indeed stimulated the commercial application of public research, no less promoted technical advance more generally. In this paper, we will exploit the 1994 Carnegie Mellon Survey (CMS) on the Nature and Determinants of Industrial R&D at least to probe some of the key assumptions of policies underpinning the privatization of public research.6 Assuming that firms need to be able to capture some significant share of the returns to their innovations to have the incentive to invest in R&D to begin with, is it reasonable to assume that patents are central to that capture? Second, should we assume that the knowledge flowing into the firm (as opposed to out of the firm) also has to be protected in some way for the firm to be willing to devote the complementary efforts required to bring it to market?
One of the themes of this paper and the prior research upon which it builds is that the answers to most of these questions will differ across industries. The CMS data allow us to address many of these issues on an industry-by-industry basis. We will exploit these data to consider the experience of four industries. We will examine computers, semiconductors and drugs. For the purpose of this analysis we win also break out our drug industry respondents into drugs and biotechnology on the basis of their self-reported R&D and market activities. Although not a distinct industry, we will consider biotechnology as such for this paper.
Section 2 briefly describes the CMS data. Section 3 provides background on the impact of public research on industrial R&D. Section 4 considers the channels through which public research impacts industrial R&D and the role that patents might play in affecting those knowledge flows. Section 5 considers the importance of the different mechanisms that firms employ to protect their inventions and, in that context, the particular role of patents. In Section 6, we examine the reasons why firms patent, and implications of those motives for the interface between public research and industry. We conclude the paper in Section 7.
DATA AND METHOD
The data come from a survey of R&D managers administered in 1994. The population sampled includes all the R&D labs or units located in the U.S. con-
ducting R&D in manufacturing industries as part of a manufacturing firm. The sample was randomly drawn from the eligible labs listed in the Directory of American Research and Technology or belonging to firms listed in Standard and Poor’s COMPUSTAT, stratified by 3-digit SIC industry.7 In our survey, we asked R&D unit or lab managers to answer questions with reference to the “focus industry” of their R&D lab or unit, defined as the principal industry for which the unit was conducting its R&D. We sampled 3,240 labs,8 and received 1,478 responses, yielding an unadjusted response rate of 46 percent and an adjusted response rate of 54 percent.9 Our survey data are supplemented with data on firm sales and employees from COMPUSTAT, Dun and Bradstreet, Moodys, Ward’s and similar sources.
We have 40 observations in our drug industry sample, 21 in biotechnology, 34 in computers, and 25 in semiconductors. The average overall firm sales for our respondents is almost $3 billion in the drug industry, $1.8 billion in biotechnology, $4.4 billion in computers, and $6.3 billion in semiconductors. The average R&D professional employment in the focus industry for each of our respondents is 577 in drugs, 234 in biotechnology, 1580 in computers and 302 in semiconductors.
IMPORTANCE OF PUBLIC RESEARCH TO INDUSTRIAL R&D
As background to our discussion of the impacts of intellectual property policy on the interface between public research and industrial R&D, it is useful to consider the importance of public research to industrial R&D more generally. As noted in Cohen et al., one widely accepted view holds that the short-term impact of university research is negligible except in a few industries.10 Accumulating evidence suggests, however, that we may wish to revise this view. Studies published since 1989, as well as the results of the 1994 Carnegie Mellon Survey, suggest that university research provides critical short-term payoffs in some in-
dustries (such as drugs), and is broadly important in numerous industries.11 A key result that should not be revised, however, is that the impacts of university research vary substantially across industries, and we see this variation across drugs, biotechnology, computers and semiconductors.
The CMS data provide a number of different measures of the impact of public research on ‘industrial R&D. For brevity, we will only focus on a couple of these. The results reported here are, however, qualitatively similar across all the measures. We begin our analysis by examining responses to a key question in the survey that considers the importance to each respondent’s R&D activities of information originating from a broad range of information sources, of which university and government R&D—collectively labeled “public research” —are one. In order to make the notion of the importance of information from these sources tangible, we asked respondents to indicate whether, over the prior three years, information from a source either suggested new R&D projects or contributed to the completion of existing projects over the prior three years. Presented in Figure 1, the results show that the majority of respondents for both drug and
biotechnology companies report that public research both suggested new R&D projects and contributed to the completion of R&D projects. Relative to the results for the full sample spanning all of manufacturing reported by Cohen et al., these two industries score among the highest.12 Respondents from the semiconductor industry evaluate the contribution of public research to be almost as high as that for drugs and biotechnology in both categories of contribution. In contrast, public research has relatively little impact on the R&D of our computer industry respondents. One point to recognize here is that the conventional view of public research is that it principally suggests new ideas, presumably spawning new projects. That is not the case. Our data indicate that in the three industries where its contribution is substantial, it contributes both by suggesting new R&D projects and by increasing the efficiency of existing R&D projects.
The same cross-industry qualitative pattern of importance of public research is observed in Figure 2 that presents the percentage of firms’ R&D projects reported to make use of the findings of public research over the prior three years. Public research was obviously a central source of information for biotechnology firms, important for drug and semiconductor firms, and considerably less important for computer firms.13
Thus, we conclude that public research is most central to the R&D activities of drug and biotechnology firms, perhaps a bit less important for the R&D of semiconductor firms (though still quite important), and least central to the R&D of computer firms.
INFORMATION CHANNELS
In this section, we examine the findings from the CMS on the importance of the different channels through which public research findings might flow to industrial R&D labs. By examining the importance of the different channels, we hope to arrive at some idea of the extent to which firms might rely on channels of information from public research institutions that are public, private or lend themselves to privatization.
We asked our respondents to report on a four point Likert scale the importance to a recently completed major R&D project of each of ten possible sources of information on the research findings or R&D activities of universities or government R&D labs or institutes. The information sources considered include patents, informal information exchange, publications and reports, public meetings and conferences, recently hired graduates, licenses, joint or cooperative ventures, contract research, consulting and temporary personnel exchanges. For each of the four industries, Figure 3 shows the percent of respondents reporting that a given source was at least “moderately important.” We observe that the most important channels across all four industries are the public ones, namely publications and reports and meetings or conferences. Informal information exchange is also quite important. The results reported here are similar to those for the manufacturing sector as a whole as reported in Cohen et al.14 There, the four dominant channels of communication between public research and industrial R&D are, in order of importance, publications and reports, public meetings and conferences, informal information channels and consulting. A factor analysis also indicated that these four information channels tend to be used together. Thus, person-to-person interactions, such as informal information exchange or consulting, tend to be used in conjunction with more public channels such as publications or conferences.15 Although we do not know the reasons for the apparent complementarity across these channels, one might imagine, for example, that firms contact faculty or engage them as consultants in response to the publication or public presentation of research judged to offer commercial potential.
The results presented in Figure 3 also reflect some clear differences in the employment of information channels across the four industries. First, a comparison of this figure with Figures 1 and 2 suggests that all or most of the channels tend to be more important to the extent that public research is more generally reported to be more relevant to each industry’s R&D. Thus, we observe that almost all the channels (except for recent hires and temporary personnel exchanges) are reported to be less important by our computer industry respondents. We also see differences in the importance of channels across the three remaining industries reporting the contribution of public research to be relatively high. First, we see a sharp difference between drug and biotechnology industry respondents versus semiconductor industry respondents. Specifically, most of the formal and market-mediated channels, including, patents (which may convey information via licenses or freely through patent disclosures), licenses, joint ventures and contracts are much less important for the semiconductor respondents than they are for the drug and biotechnology respondents. The importance of communication channels differ even between drugs and biotechnology. For the former, consulting is far more important. For biotechnology, public meetings and conferences are more important, as are licenses to some extent.16
Our results offer several implications and raise numerous questions. First, the dominant importance of public channels for disseminating public research across all four industries suggest that public research findings disseminated via the standard channels of “open science” are commonly used by industrial R&D labs. In areas, however, where public research has greater impacts such as drugs or biotechnology, channels that are private and more restricted, such as consulting, contract research and licenses, appear to be more important. Even where, however, such channels are important, observe that the public channels are more important still.
Although we cannot yet explain the cross-industry differences in the relative importance of the different channels, it is possible that market mediated channels, notably licenses, contracts and consulting, are used more intensively to the extent that patents offer more protection, which, as shown below, they do in the biotechnology and drug industries. This raises the question of the extent to which the overall impact of public research on industrial R&D depends on the effectiveness of specific channels, and particularly private channels. A different possibility—and one that is not particularly consistent with the rationale for Bayh-Dole-is that the overall impact of public research on industrial R&D may have little to do with the strength of any specific or set of channels, but rather simply
reflects the overall relevance of public research. In this view, if the research is relevant and useful to industry, it will be used somehow. There is some support for this position in that the three industries reporting public research to be important overall (per Figures 1 and 2) rate most of the channels to be more important than does the one industry, computers, reporting public research to be relatively unimportant.
APPROPRIABILITY MECHANISMS
In this section, we review the CMS results on how firms in the four industries protect their innovations, and address the question of the role of patents.17 For the sake of brevity, we focus on the effectiveness of the different mechanisms that firms use to appropriate the returns to their product innovations, including patents, secrecy, lead time, complementary sales and service and complementary manufacturing facilities and know how.18 The role of these mechanisms in protecting process innovations is reported in Cohen et al.19
To measure the effectiveness of appropriability mechanisms, we asked respondents to report the percentage of their product and process innovations for which each appropriability mechanism had been effective in protecting their firm’s competitive advantage from those innovations during the prior 3 years. The five response categories were: 1) less than 10 percent, 2) 10 percent through 40 percent, 3) 41 percent through 60 percent, 4) 61 percent through 90 percent, and 5) greater than 90 percent. This response scale reflects how central a mechanism is to firms’ strategies of appropriating rents to their innovations in the sense that it reflects both the frequency with which a mechanism is employed and the effectiveness of that mechanism given its use, and we interpret effectiveness as reflecting the ability of a mechanism to protect the profits accruing to the commercialization or licensing of the protected invention.20
For each of the four industries, Figure 4 presents the percentage of respondents reporting a given mechanism to beat least “moderately effective.” Figure 4 shows that patents are reported to be among the most effective mechanisms for
drug and biotechnology firms. For these two industries, secrecy is comparably important to patents, and lead time is almost as important as patents and secrecy for drug companies. In contrast to the drug and biotechnology industries, we observe that patents are not among the key means used to protect innovations in either the computer or semiconductor industries. In those two industries, firms rely more heavily on secrecy, lead time and complementary capabilities to protect their inventions. A comparison of these results to those for the manufacturing sector as a whole reported in Cohen et al. suggests that the semiconductor and computer industries are not unusual.21 Similar to the earlier findings of Scherer et al., Mansfield et al., and Levin et al., patents are reported to be the least effective of the various means employed by firms to protect their inventions in the preponderance of manufacturing industries, with secrecy, lead time and the use of complementary capabilities dominant.22 This implies that the high effectiveness score of patents observed in the drug and biotechnology industries is unusual. In fact, in Cohen et al., where biotechnology is included in the drug industry category, patents are reported to be more effective in drugs than in any other industry.23
Although patents are not considered to be among the most effective mechanisms for protecting the profits due to the commercialization or licensing of invention in the computer or semiconductor industries—or for manufacturing firms in general—that does not mean that they are not used or are inessential. Cohen et al.’s analysis of the relationships across the appropriability mechanism effectiveness scores for the manufacturing sector as a whole indicates that appropriability mechanisms are not mutually exclusive in the ways that they are employed.24 Firms will often use more than one to protect even the same invention. In drugs, for example, firms may well use a combination of secrecy, patents and lead time advantage to protect the same product. Thus, to say that patents are less effective relative to other mechanisms does not mean they do not confer value. Rather, it suggests that they are less central to the methods firms use to protect their inventions.
Aside from firms’ judgments of patent effectiveness, the CMS data also provide estimates of the relative frequency with which firms apply for patents. CMS data include measures of patent application propensities, defined as the percentage of each respondent’s innovations for which they apply for a patent. Figure 5 presents the average product (and process) patent propensities by indus-
try. The results sensibly suggest that the industry, namely biotechnology, which rates patent effectiveness the highest has the highest patent propensity. It is surprising, however, that computer industry respondents that rate patents to be less effective than drug industry respondents, report an average patent propensity of 37 percent, which is higher than the average propensity of 28 percent reported by drug industry respondents. This result changes, however, once we focus on more R&D active firms. For drug industry respondents with greater than the median R&D intensity, product patent propensity is 47 percent as compared to a comparably computed 40 percent for the computer industry. The average patent propensity for the semiconductor industry is the lowest of the four industries, which is consistent with their low rating of patent effectiveness. Thus, product patent propensities among more R&D intensive firms tend to be higher in industries where patents are reported to be more effective.
These results lead us to the question of why firms do not patent; what are the limits on their effectiveness? The CMS data provide some insight here. In our survey, we asked firms to report the reasons contributing to their most recent decision not to apply for a patent. The reasons for not applying for a patent considered in our survey include: 1) Difficulty in demonstrating the novelty of an invention; 2) the amount of information disclosed in a patent application; 3) the cost of applying; 4) the cost of defending a patent in court; 5) the ease of legally inventing around a patent. Figure 6 indicates the percentage of respon-
dents indicating whether a particular reason figured into their decision not to patent.25 Our results in Figure 6 show that ease of inventing around is the most cited reason, and is of greater concern to the more R&D active firms in drugs and biotechnology than in computers and semiconductors. Aside from concerns over the demonstration of novelty—and hence whether an invention is patentable to begin with—the disclosure of information that comes with patenting is also of concern. The only cross-industry differences of note is that concern over the disclosure of information appears to be of less concern to the semiconductor respondents than others, perhaps because the technology is sufficiently complex and know-how based such that a patent discloses less of real value to rivals. Also, concern over the cost of either the patent application or subsequent legal defense is less of an impediment to patenting in the drug industry.26
REASONS TO PATENT
In this section, we will try to develop a better understanding of how firms use patents, particularly when patenting does not appear to protect the commercialization or licensing of their patented inventions. We will also speculate on the implications of the uses of patents for the exploitation of public research.
In our survey, we asked respondents to indicate which of seven possible reasons motivated their most recent decisions to apply for a patent. The reasons for patenting considered in our survey include the prevention of copying, the prevention of other firm’s attempts to patent a related invention (which we call “patent blocking”), the earning of licensing revenue, use to strengthen the firm’s position in negotiations with other firms (as in cross-licensing agreements), the prevention of infringement suits, use as a measure of internal performance of a firms’ technologists, and the enhancement of the firm’s reputation. Figure 7 presents the percent of respondents by industry indicating a given reason applied to their most recent decision to patent.
Aside from the prevention of copying, the most prominent reasons for patenting across all four industries include patent blocking, prevention of suits and for use in negotiations. This result resembles that for the manufacturing sector as a whole reported in Cohen et al.27 One cross-industry difference observed in
Figure 7 is that licensing revenue is a less pervasive motive for patenting among firms in the computer industry, but using patents for negotiations is somewhat more pervasive, implying that most of those negotiations are associated with cross-licensing. Patent blocking appears to be more pervasive in biotechnology and drugs than in computers or semiconductors, though high in all four industries. Patenting to enhance the firm’s reputation is less evident in the semiconductor and computer industries than in drugs or biotechnology.
In the more comprehensive analysis of the reasons to patent for the manufacturing sector as a whole, Cohen et al. highlight other cross-industry differences in the reasons to patent.28 They find that once one focuses on firms that patent more aggressively (by weighting responses by each respondent’s number of patent applications), a sharp difference becomes apparent between semiconductor and drug industry respondents (which include biotechnology firms in the Cohen et al analysis).29 Firms in the drug industry appear to use patents in the way they are typically thought to be used, namely to protect the commercialization or licensing of the patented inventions. Firms also use patents to block the patenting of substitute or competing inventions, but rarely use patents for crosslicensing negotiations. In contrast, in semiconductors, the same patents that are used for blocking are also commonly used for strengthening the firm’s position in crosslicensing negotiations, suggesting that patent blocking tends to be directed against not substitute but complementary inventions. This makes sense in a complex product industry like semiconductors where commercialized innovations are often comprised of numerous separately patentable elements and incumbent manufacturing firms consequently either require access to one another’s technology, or at least the freedom to pursue projects that are similar to those of rivals without being sued.30 Indeed, on the basis of both the survey data and field interviews, Cohen et al.31 conclude that firms will often patent in semiconductors not to profit directly from a particular patented invention through either its commercialization or licensing, but to build portfolios to compel their inclusion in cross-licensing negotiations or at least provide some protection against suits by other manufacturers.32
The use of patent portfolios to achieve such “player” status in the semiconductor industry suggests that patents will tend to be most useful to firms that can generate or acquire rights over numerous patents and have access to the considerable legal resources essential to such a strategy. Where firms are both active in
the manufacture of semiconductors and conduct R&D, they will tend to cross-license among themselves in recognition of their mutual dependence rather than extract licensing revenue. The example of Texas Instruments (T.I.) suggests, however, that where firms do not hold a large stake in manufacturing the product but nonetheless possess patents on key technology, there is little benefit to be realized from cooperative cross-licensing agreements. Rather, since firms in T.I.’s position do not depend on access to other firms’ technologies and thus cannot be hurt al. 1 that much by a countersuit, they have a lot to gain and risk losing little by aggressively defending their patents and pursuing licensing revenue.33
Use of patents in the computer industry appears to differ from their use in either the drug or semiconductor industry, and has some similarities to both. First, patenting for use in cross-licensing negotiations is common in the computer industry, and, to this extent, is similar to the practice in the semiconductor industry. Yet, using these same patents to block appears to be less common. Patenting to earn licensing revenue is more pervasive in the computer industry among the larger, more patent intensive firms.
The contrast between the motives for patenting among drug and biotechnology firms, on the one hand, and particularly semiconductor firms on the other, challenges the rationale behind Bayh-Dole and similar policies. For drug and biotechnology firms, patents indeed appear to play a role in stimulating the commercialization of invention, as Bayh-Dole assumes. On the other hand, in semiconductors, patents play a different role. They are not used by firms to protect the commercialization of invention. Rather, the accumulation of patents into large portfolios confers player status in the industry, and, in turn enables access to the technology of rivals and the ability to pursue R&D and commercialization without fear of suits from other incumbents. In this latter setting, what is the use of a patent on an invention originating from public research laboratories? It is unlikely that such patents provide protection that induces the follow-on R&D and other investments necessary for commercialization. Thus, patents on public research in areas such as semiconductors may be of limited relevance to the enterprise of commercialization, and consequently play little of the role envisioned by the framers of the Bayh-Dole Amendment.
It does appear, however, that patents on public research outputs have stimulated universities and other publicly funded institutions to attempt to garner licensing revenues from firms. Although benefitting universities, such actions also raise the cost of innovation and may dampen firms’ incentives to conduct follow-on complementary R&D in the absence of offsetting benefits. And, from the discussion above, we conjecture that the offsetting benefits may be rather slight in industries such as semiconductors. There is another potentially important but
subtle disadvantage to such rent-seeking behavior on the part of public research institutions. Where the progress of public research itself depends upon reciprocal information flows between industry and public research institutions, the prospect of, say, a university’s staking of exclusive rights may undermine that progress by chilling those flows. Firms may be reluctant to communicate freely about their own technology with public research institutions if they believe that technology may end up in a university patent for which they will have to pay.34
DISSCUSSION
Our data do not suggest that the privatization of public research flows into industry is essential to securing the commercial application of much of that information. Although we do observe a substantial impact of public research in drugs and biotechnology where patents and associated information channels are strong, we also observe a substantial impact in semiconductors where that is not the case. Moreover, across all four industries that we have examined, the most important channels from public research to industrial R&D labs are the most public ones. Moreover, our analyses suggest that even when private channels such as consulting are strong, use of public channels complement that strength.
The cross-industry differences observed in the impact of public research on industrial R&D, the importance of the different channels of communication, and the importance and role of patents in particular also offer other implications. First, public research simply matters more in some industries than others, suggesting that to the extent that Bayh-Dole and related policies have an impact, that impact will vary substantially across industries. For example, we observe that public research appears to be particularly important for three of the four industries we examined, namely drugs, biotechnology and semiconductors. So, in these settings, what is at stake may be quite large. In these three industries, we also fmd, however, important differences in the effectiveness of patents in protecting the commercialization and licensing of inventions. Thus, in drugs and biotechnology where firms report patents to be quite effective, patents on the outputs of public research that are already close to commercialization may play a role in stimulating downstream commercialization. The enforcement of exclusionary rights on more “upstream” public research outputs essential to the research of a broader community may also impose, however, considerable costs, as noted by Mazzoleni and Nelson.35
In contrast to the role and importance of patents in drugs and biotechnology, patents are not reported to be very effective in protecting the commercialization
of innovation in the semiconductor industry. Here, firms protect their inventions through a combination of lead time advantages, secrecy and the exploitation of complementary capabilities. Firms in semiconductors do, however, patent their inventions, but apparently a key motive, at least among larger incumbents, is to amass portfolios used to gain access to other firms’ technologies via cross-licensing and to protect themselves from suits. It is not apparent how patents on public research bearing on semiconductor technology would therefore foster subsequent commercialization, which is the rationale for Bayh-Dole. If granted exclusive licenses, semiconductor firms could conceivably use patents originating from public research like their own that is to strengthen their bargaining position in the industry.36 There is also, however, the prospect that if public research institutions attempt to enforce their property rights, they could antagonize firms and undermine the reciprocal communication flows that contribute to the research efforts of all parties.
REFERENCES
Barton, John (1999) “Reforming the Patent System,” Working paper, Stanford Law School.
Cohen, W.M., Florida, R., Randazzese, L., Walsh, J. (1998), “Industry and the Academy: Uneasy Partners in the Cause of Technological Advance,” in Noll, R., ed., Challenges to Research Universities, Brookings Institution.
Cohen, W.M., Nelson, R.R. and Walsh, J. (2000) “Protecting Their Intellectual Assets: Appropriability Conditions and Why U.S. Manufacturing Firms Patent (or Not)” National Bureau of Economic Research, Working Paper No. 7552.
Faulkner, W. and Senker, J. (1995) Knowledge Frontiers: Public Sector Research and Industrial Innovation in Biotechnology, Engineering Ceramics, and Parallel Computing, New York: Oxford University Press.
Gibbons, M. and Johnston, R. (1975) “The Roles of Science in Technological Innovation,” Research Policy, vol. 3, pp. 220–242.
Hall, B. and Ham, R.M. (1998) “The Patent Paradox Revisited: Firm Strategy and Patenting in the U.S. Semiconductor Industry,” NBER Conference on Patent System and Innovation, Jan. 8, 9, Santa Barbara, 1999, Working Paper.
Heller, M. and Eisenberg, R. (1998) “Can Patents Deter Innovation? The Anticommons in Biomedical Research”, Science, Vol. 28, May 1, pp. 698–701.
Klevorick, A.K., Levin, R., Nelson, R.R. and Winter, S. (1994) “On the Sources and Significance of Interindustry Differences in Technological Opportunities,” Research Policy, vol. 24, no. 2, pp. 195–206.
Levin, R., Klevorick, A., Nelson, R.R., and Winter, S.G. (1987) “Appropriating the Returns from Idustrial R&D”, Brookings Papers on Economic Activity, pp. 783–820.
Mansfield, E. (1986) “Patents and Innovation: An Empirical Study,” Management Science, 32:173–181.
Mansfield, E., Schwartz, M. and Wagner, S. (1981) “Imitation Costs and Patents: An Empirical Study,” Economic Journal, 91:907–918.
Mazzoleni, R. and Nelson, R.R. (1998), “Economic theories about the benefits and costs of patents,” Journal of Economic Issues, 32(4):1031–1052.
Mowery, D.C., Nelson, R.R., Sampat, B.N., and Ziedonis, A.A., “The Effects of the Bayh-Dole Act on U.S. University Research and Technology Transfer: An Analysis of Data from Columbia University, the University of California and Stanford University,” 1999.
Narin, F., Hamilton, Kimberly S., and Olivastro, D. (1997) “The Increasing Link between U.S. Technology and Public Science,” Research Policy, 26(3):317–330.
Rosenberg, N. and Nelson, R.R. (1994) “American Universities and Technical Advance in Industry,” Research Policy, 239(3):323–348.
Scherer, F.M., et al. (1959) Patents and the Corporation. 2nd ed. Boston, privately published.