Collaborative R&D: How Effective Is It?

R&D collaboration is widespread in the U.S. economy of the 1990s. Literally hundreds of agreements now link the R&D efforts of U.S. firms, and other collaborative agreements involve both U.S. and non-U.S. firms. Collaboration between U.S. universities and industry also has grown significantly since the early 1980s-hundreds of industry-university research centers have been established, and industry’s share of U.S. university research funding has doubled during this period, albeit to a relatively modest 7 percent. Collaboration between industrial firms and the U.S. national laboratories has grown as well during this period, with the negotiation of hundreds of agreements for cooperative R&D.

R&D collaboration has been widely touted as a new phenomenon and a potent means to enhance economic returns from public R&D programs and improve U.S. industrial competitiveness. In fact, collaborative R&D projects have a long history in U.S. science and technology policy. Major collaborative initiatives in pharmaceuticals manufacture, petrochemicals, synthetic rubber, and atomic weapons were launched during World War II, and the National Advisory Committee on Aeronautics, founded in 1915 and absorbed into NASA in 1958, made important contributions to commercial and military aircraft design throughout its existence. Similarly, university-industry research collaboration was well established in the U.S. economy of the 1920s and 1930s and contributed to the development of the academic discipline of chemical engineering, transforming the U.S. chemicals industry.

There is no doubt that collaborative R&D has made and will continue to make important contributions to the technological and economic well-being of U.S. citizens. But in considering the roles and contributions of collaboration, we must focus on the objectives of collaborative programs, rather than treating R&D collaboration as a “good thing” in and of itself. Collaborative R&D can yield positive payoffs, but it is not without risks. Moreover, R&D collaboration covers a diverse array programs, projects, and institutional actors. No single recipe for project design, program policies, or evaluation applies to all of these disparate entities.

In short, R&D collaboration is a means, not an end. Moreover, the dearth of systematic analysis and evaluation of existing federal policies toward collaboration hampers efforts to match the design of collaborative programs to the needs of different firms, industries, or sectors. A review of U.S. experience reveals a number of useful lessons and highlights several areas where more study is needed.

Policy evolution

Since the mid-1970s, federal policy has encouraged collaboration among many different institutional actors in the U.S. R&D system. One of the earliest initiatives in this area was the University-Industry Cooperative Research program of the National Science Foundation (NSF), which began in the 1970s to provide partial funding to university research programs enlisting industrial firms as participants in collaborative research activities. The NSF efforts were expanded during the 1980s to support the creation of Engineering Research Centers, and other NSF programs now encourage financial contributions from industry as a condition for awarding research funds to academic institutions. Moreover, the NSF model has been emulated by other federal agencies in requiring greater cost-sharing from institutional or industry sources in competitive research grant programs. The NSF and other federal initiatives were associated with the establishment of more than 500 university-industry research centers during the 1980s.

R&D collaboration between industrial firms and universities received another impetus from the Bayh-Dole Act, passed in 1980 and amended in 1986, which rationalized and simplified federal policy toward the patenting and licensing by nonprofit institutions of the results of publicly funded research. The Bayh-Dole Act has been credited with significant expansion in the number of universities operating offices to support the patenting, licensing, and transfer to industrial firms of university research results. These offices and the legislation have also provided incentives for industrial firms to form collaborative R&D relationships with universities.

The Bayh-Dole Act, the Stevenson-Wydler Act of 1980, and the Technology Transfer Act of 1986 also created new mechanisms for R&D collaboration between industrial firms and federal laboratories through the mechanism of the Cooperative Research and Development Agreement (CRADA). Under the terms of a CRADA, federal laboratories are empowered to cooperate in R&D with private firms and may assign private firms the rights to any intellectual property resulting from the joint work; the federal government retains a nonexclusive license to the intellectual property. The XXXXXX XXXXXX Act [Which Act?] was amended in 1989 to allow contractor-operated federal laboratories to participate in CRADAs. Federal agencies and research laboratories have signed hundreds of CRADAs since the late 1980s; between 1989 and 1995, the Department of Energy (DOE) alone signed more than 1,000 CRADAs. The 1996 Technology Transfer Improvements and Advancement Act strengthened the rights of industrial firms to exclusively license patents resulting from CRADAs.

Federal antitrust policy toward collaborative R&D also was revised considerably during the early 1980s. Through much of the 1960s and 1970s, federal antitrust policy was hostile toward R&D collaboration among industrial firms. The Carter administration’s review of federal policies toward industrial innovation resulted in a new enforcement posture by the Justice Department, embodied in guidelines issued in 1980 that were less hostile toward such collaboration. In 1984, the passage of the National Cooperative Research Act (NCRA) created a statutory “safe harbor” from treble damages in private antitrust suits for firms registering their collaborative ventures with the Justice Department. The NCRA was amended to incorporate collaborative ventures in production in 1993. During the period from 1985 through 1994, U.S. firms formed 575 “research joint ventures,” the majority of which focused on process R&D. Interestingly, Justice Department data on filings under the NCRA since the passage of the 1993 amendments report the formation of only three joint production ventures.

Finally, the federal government began under the Reagan administration to provide financial support to R&D consortia in selected technologies and industries. The most celebrated example of this policy shift is SEMATECH, the semiconductor industry R&D consortium established in 1987 with funding from the federal government (until 1996), industry, and the state of Texas. Since 1987, the Advanced Technology Program, established under the Bush administration, has provided matching funds for a number of industry-led R&D consortia, some of which involve universities or federal laboratories as participants. More recent programs such as the Technology Reinvestment Program and the Project for a New Generation of Vehicles have drawn on funding from other federal agencies to supplement industry financial contributions for the support of industry-led R&D consortia.

Although the federal policy has shifted dramatically in the past 20 years and spawned a diverse array of collaborative arrangements, surprisingly little effort has been devoted to evaluation of any one of the legislative or administrative initiatives noted above. For example, how should one interpret the evidence on the small number of production joint ventures filed with the Justice Department since 1993? A broader assessment of the consistency and effects of these policies as a whole is needed. Recognizing the number of such initiatives implemented in a relatively short period of time, their occasionally inconsistent structure, and their potentially far-reaching effects, this comprehensive assessment should precede additional legislation or other policy initiatives.

Benefits and risks

A brief discussion of the potential benefits and risks of R&D collaboration is useful to assess the design and implementation of specific collaborative programs. The economics literature identifies three broad classes of benefits from R&D collaboration among industrial firms: (1) enabling member firms to capture “knowledge spillovers” that otherwise are lost to the firm investing in the R&D that gives rise to them, (2) reducing duplication among member firms’ R&D investments, and (3) supporting the exploitation of scale economies in R&D. This group of (theoretical) benefits has been supplemented by others in more recent discussions of policy that often address other forms of collaboration: (1) accelerating the commercialization of new technologies, (2) facilitating and accelerating the transfer of research results from universities or public laboratories to industry, (3) supporting access by industrial firms to the R&D capabilities of federal research facilities, and (4) supporting the creation of a common technological “vision” within an industry that can guide R&D and related investments by public and private entities.

This is a long list of goals for any policy instrument. Moreover, many of these goals deal with issues of technology development and commercialization rather than scientific research. Although a sharp separation between scientific research and technology development is unwarranted on empirical and conceptual grounds, the fact remains that collaboration in “R” raises different issues and poses different challenges than does collaboration in “D” or in R&D.

The benefits of collaborative R&D that economists have cited in theoretical work are difficult to measure. More important, however, they imply guidelines for the design of R&D collaboration that may conflict with other goals of public R&D policy. The hypothesized ability of industry-led consortia to internalize knowledge spillovers, for example, is one reason to expect them to support more fundamental, long-range research. Nonetheless, most industry-led consortia, including SEMATECH, support R&D with a relatively short time horizon of three to five years. In addition, most industry-led R&D consortia seek to protect jointly created intellectual property. Yet protection of the results of collaborative R&D may limit the broader diffusion and exploitation of these results that would increase the social returns from these investments. When industry-led consortia receive public financial support, this dilemma is sharper still.

A similar tension may appear in collaborations between U.S. universities and industrial firms, especially those centered around the licensing of university research results. In fact, university research has long been transferred to industrial enterprises through a large number of mechanisms, including the training of graduates, publication of scientific papers, faculty consulting, and faculty-founded startup firms. Efforts by universities to obtain strong formal protection of this intellectual property or restrictive licensing terms may reduce knowledge transfer from the university, with potentially serious economic consequences. There is no compelling evidence of such effects as yet, but detailed study of this issue has only begun.

Reduced duplication among the R&D strategies of member firms in consortia and other forms of R&D collaboration is another theoretical benefit that may be overstated. The experience of participants in industry-led consortia, collaborations between federal laboratories and industry, and university-industry collaborations all suggest that some intrafirm R&D investment is essential if the results of the R&D performed in the collaborative venue are to be absorbed and applied by participating firms. In other words, some level of in-house duplication of the R&D performed externally is necessary to realize the returns from collaborative R&D.

The other goals of R&D collaboration that are noted above raise difficult issues. For example, the reduction of duplicative R&D programs within collaborating firms and the development by an industry of a common technological vision both imply some reduction in the diversity of scientific or technological avenues explored by research performers. Since one of the hallmarks of technology development, especially in its earliest stages, is pervasive uncertainty about future developments, the elimination of such diversity introduces some risk of collective myopia. One may overlook promising avenues for future research or even bypass opportunities for commercial technology development. A single-minded industry vision can conserve resources, but it may be risky or even ill-advised when one is in the earliest stages of development of a new area of science or technology. After all, the postwar United States has been effective in spawning new technology-intensive industries precisely because of the ability of the U.S. market and financial system to support the exploration of many competing, and often conflicting, views of the likely future path of development of breakthroughs such as the integrated circuit, the laser, or recombinant DNA techniques.

Managing R&D collaboration between industrial firms and universities or federal laboratories is difficult, and problems of implementation and management frequently hamper the realization of other goals of such collaboration. Collaborative R&D may accelerate the transfer of research results from these public R&D performers to industry, but the devil is in the details. The sheer complexity of the management requirements for R&D collaborations, especially those involving many firms and more than one university or laboratory, may slow technology transfer. In addition, the costs of such transfer-including the maintenance by participating firms of parallel R&D efforts in-house and/or the rotation of staff to an offsite R&D facility-may exceed the resources of smaller firms. In some cases, the effectiveness of CRADAs between federal laboratories and university-industry collaborations has been impeded by negotiations over intellectual property rights, regardless of the actual importance of such rights, in order to conform with the statutory and administrative requirements of such collaborations.

A beginning at differentiation

At the risk of oversimplifying a very complex phenomenon, one can single out three categories of R&D collaboration as especially important: (1) industry-led consortia, which may or may not receive public funds; (2) collaborations between universities and industry; and (3) collaborations between industry and federal laboratories, often supported through CRADAs. These forms of collaboration have received direct encouragement, and in some cases financial support, from federal policy in the past 20 years. In addition to the variety of collaborative mechanisms, there is considerable variation among technology classes in the types of policies or organizational structures that will support effective R&D performance and dissemination.

Industry-led consortia. As noted earlier, these undertakings rarely focus on long-range research. Indeed, many consortia in the United States pursue activities that more closely resemble technology adoption than technology creation. SEMATECH, for example, has devoted considerable effort to the development of performance standards for new manufacturing equipment. These efforts are hardly long-range R&D, but they can aid equipment firms’ sales of these products and SEMATECH members’ adoption of new manufacturing technologies. Industry consortia also do not eliminate duplication in the R&D programs of participants because of the requirements for in-house investments in R&D and related activities to support inward transfer and application of collaborative R&D results. The need for these investments means that small firms may find it difficult to exploit the results of consortia R&D, and particular attention must be devoted to their needs. Consortia may aid in the formation of an industry-wide vision of future directions for technological innovation, but such consensus views are not always reliable, especially when technologies are relatively immature and the direction of their future development highly uncertain. Such visions can be overtaken by unexpected scientific or technological developments.

Some of these features of “best practice” that have been identified with the SEMATECH experience, especially the need for flexibility in agenda-setting and adaptation, may be difficult to reconcile with the requirements of public oversight and evaluation of publicly funded programs. Moreover, the SEMATECH experience suggests that collaborative R&D alone is insufficient to overcome weaknesses in manufacturing quality, marketing, or other aspects of management. Indeed, in its efforts to strengthen smaller equipment suppliers, SEMATECH supplemented R&D with outreach and education (mainly in the equipment and materials industries) in areas such as quality management and financial management.

University-industry collaborations. Collaborative research involving industry and universities has a long history. A combination of growing R&D costs within academia and industry, along with the supportive federal legislation and policy shifts described above, have given considerable impetus to university-industry collaboration during the past 20 years. Industry now accounts for roughly 7 percent of academic R&D spending in the United States, the number of university-industry research centers has grown, and university patenting and licensing have expanded significantly since 1980. As in the case of SEMATECH, recent experience supports several observations about the effectiveness of these collaborations for industrial, academic, and national goals and welfare:

Little evidence is available about the ability of these collaborative R&D ventures to support long-term research. Cohen et al. (1994) found that most university-industry engineering research centers tended to focus on relatively near-term research problems and issues faced by industry. Other undertakings, however, such as the MARCO initiative sponsored by SEMATECH, are intended to underwrite long-range R&D efforts. University-industry collaboration thus may be able to support long-range R&D more effectively than industry-led consortia.

Preliminary evidence indicates that the Bayh-Dole Act has had little effect on the characteristics of the invention disclosures from faculty. Bayh-Dole did cause many other universities to enter into patenting and licensing activities. In addition, data from the University of California, which was active in patenting and licensing before the passage of the bill, suggest that the number of annual invention disclosures began to grow more rapidly and shifted to include a larger proportion of biomedical inventions before, rather than after, the passage of this law. These findings are preliminary, however, and a broader evaluation of the effects of the Bayh-Dole Act is long overdue.

Effective industry-university relationships differ considerably among different industries, academic disciplines, and research areas. In biomedical research, for example, individual patents have considerable strength and therefore potentially great commercial value. Licensing relationships covering intellectual property “deliverables” thus have been quite effective. In other areas, however, such as chemical engineering or semiconductors, the goals of industry-university collaborations, and the vehicles that are best suited to their support, differ considerably. Firms in these industries often are less concerned with obtaining title to specific pieces of intellectual property than with seeking “windows” on new developments at the scientific frontier and access to high-quality graduates (who are themselves effective vehicles for the transfer of academic research results to industry). For firms with these objectives, extensive requirements for specification and negotiation of the disposition of intellectual property rights from collaborative research may impede such collaboration. The design of university-industry relationships should be responsive to such differences among fields of research.

Excessive emphasis on the protection by universities of the intellectual property resulting from collaborative ventures, especially when combined with restrictive licensing terms, may have a chilling effect on other channels of transfer, restricting the diffusion of research results and conceivably reducing the social returns from university research. Unbalanced policies, such as restrictions on publication, raise particular dangers for graduate education, which is a central mission of the modern university and an important channel for university-industry interaction and technology transfer.

Management of industry-university relationships should be informed by more realistic expectations among both industry executives and university administrators on means and ends. In many cases, universities may be better advised to focus their management of such relationships and any associated intellectual property on the establishment or strengthening of research relationships, rather than attempting to maximize licensing and royalty income.

As is true of industry-led consortia, industrial participants in collaborative R&D projects with universities must invest in mechanisms to support the inward transfer and absorption of R&D results. The requirements for such absorptive capacity mean that university-industry collaborations may prove less beneficial or feasible for small firms with insufficient internal resources to undertake such investments.

Collaborations between federal laboratories and industry. Our recent examination of a small sample of CRADAs between a large DOE nuclear weapons laboratory and a diverse group of industrial firms suggests the following preliminary observations concerning this type of R&D collaboration:

Cultural differences matter. All of the firms participating in these CRADAs agreed that this DOE laboratory had unique capabilities, facilities, and equipment that in many cases could not be duplicated elsewhere. Nevertheless, their contrasting backgrounds meant that laboratory and firm researchers had different approaches to project management that occasionally conflicted. Moreover, the limited familiarity of many laboratory personnel with the needs of potential commercial users of these firms’ technologies meant that collaboration in areas distant from the laboratory’s historic missions was more difficult and often less successful.

The focus of many CRADAs on specification of intellectual property (IP) rights often served as an obstacle to the timely negotiation of the terms of these ventures. In a majority of the cases we reviewed, the participating firms were not particularly interested in patenting the results of their projects. The importance of formal IP rights differs among technological fields, but the emphasis in many CRADAs on intellectual property rights may be misplaced, and alternative vehicles for collaboration may be better suited to the support of such collaboration. As with university-industry collaboration, no single instrument will serve to support collaboration in all technologies or research fields. Laboratory and firm management needs to devote more effort to selecting projects for collaboration and must improve the fit between the project and the specific vehicle for such collaboration.

Most of the CRADAs reviewed in our study were concerned with near-term R&D or technology development. Participating firms frequently found it difficult to manage the transition from development to production without some continuing support from DOE personnel. Yet the terms of many of these CRADAs made a more gradual handoff very difficult.

As in other types of R&D collaboration, significant investments by participating firms to support inward transfer and application of the results of CRADAs were indispensable. Firms that found CRADAs to be especially beneficial had invested heavily in this relationship, including significant personnel rotation, travel, and communications. Along with the small size of their budgets, the costs of these investments made CRADAs involving small firms difficult to manage.

Who pays?

The case for public funding of collaborative R&D resembles the case for public funding of R&D more generally. This case is strongest where there is a high social return from collaborative R&D activities, and the gap between private and social returns is such that without public funding, the work would not be undertaken. But these arguments for public funding of collaborative R&D raise two important challenges to the design of such projects:

What is the appropriate “match” between public and private funding? A matching requirement creates incentives for participating firms to minimize costs and apply the results of such R&D. Setting a matching requirement at a very low share of total program costs may weaken such incentives and result in R&D that is of little relevance to an industry’s competitive challenges. However, if the private matching requirement is set at a relatively high level (for example, above 75 percent of total program costs), firms may choose not to participate in collaborative R&D or will undertake projects that would have been launched in any event. The ideal matching requirement will balance these competing objectives, but there is little guidance from economic theory or prior experience to inform such a choice.

If R&D collaboration seeks to encourage research investments yielding high social returns, the case for tight controls on the dissemination of the results of such R&D is weak. The assignment to private firms of intellectual property rights to the results of such R&D that is allowed by the Bayh-Dole Act and other policies is intended to encourage the commercialization of these results by establishing a stronger reason for their owners to undertake such investments. But by limiting the access of other firms to these results, patents or restrictive licenses may slow the diffusion of R&D results, reducing the social returns from the publicly funded R&D. This dilemma is another one for which neither economic theory nor program experience provides much guidance. As a general rule, however, broad patents and restrictive licensing terms for patents resulting from publicly funded collaborative R&D should be discouraged. This policy recommendation suggests that the competitive effects of any greater tilt toward exclusivity in the licensing of these patents, such as that embodied in the Technology Transfer Improvements and Advancement Act, should be monitored carefully.

These dilemmas apply to public funding of R&D, especially civilian R&D performed within industry, regardless of whether R&D collaboration is involved. The mere presence of a collaborative relationship does not eliminate them, and in some cases may complicate their resolution.

The “taxonomy” of R&D collaborations discussed earlier is hardly exhaustive, but it suggests the need for a clearer assessment of the links between the goals of R&D collaborations and their design. For example, R&D collaborations established to support long-range R&D may be more effective if they link universities and industry, rather than being undertaken through industry-led consortia. At the same time, the effects of collaboration on the other missions of U.S. universities must be monitored carefully so as not to undercut performance in these areas. Small firms often face serious problems with R&D collaboration, because of the significant investments that participants must make in technology absorption and the inability of R&D collaboration to upgrade technological capabilities in firms lacking them. In addition, small firms often need much more than technological assistance alone in order to improve their competitive performance. R&D collaborations that seek to accelerate technology access and transfer must be designed to avoid administrative requirements that may instead slow these activities. In particular, negotiations over intellectual property rights must be handled flexibly and in a manner that is responsive to the needs of all the participants.

The variations among different types of R&D collaboration are substantial, and policymakers and managers alike should proceed with great caution in reaching sweeping conclusions or in developing detailed policies that seek to govern collaboration in all institutional venues, technologies, and industries. Broad guidelines are appropriate and consistent with Congress’s role in ensuring that these undertakings serve the public interest. But the implementation of these guidelines and detailed policies governing R&D collaborations are best left to the agencies and institutions directly concerned with this activity. Greater flexibility for federal agencies in negotiating the terms of CRADAs within relatively broad guidelines, for example, would facilitate their more effective use and more careful consideration of alternatives to these instruments for collaboration.

The phenomenon of R&D collaboration has grown so rapidly that hard facts and robust generalizations about best practice and policy are exceedingly difficult to develop for all circumstances. A more comprehensive effort to collect data on R&D collaboration, perhaps spearheaded by the Commerce Department’s Technology Administration, and greater efforts to capture and learn from the results of such ventures surely are one of the most urgent prerequisites to any effort to formulate a broader policy on R&D collaboration.