The final version of this paper appears as chaper 8 of The Oxford Handbook of Innovation, eds. Jan Fagerberg, Mavid C. Mowery and Richard R. Nelson (Oxford University Press, 2005), pp 209-239.
This article offers convincing skeptical commentary on both the economic impacts of Bayh-Dole based technology transfer and on its relevance to other countries’ innovation systems. Although it does not cover the 2005-2010 period, its criticisms are relevant to the National Nanotechnology Initiative, which is a linear decendent of the Bayh-Dole model.
The piece starts with an understanding of the Bayh-Dole Act (BD) as the fulfillment of the “linear model” of innovation developed by Vannevar Bush in 1945 (Science: The Endless Frontier). BD sought to accelerate the development of commercial applications by improving the transfer of technology from academic laboratories to industrial product development: the mechanism was the university or non-profit’s new ability to own intellectual property created by its employees, and to license that IP to commercial firms that would bring the invention to market. The more pervasive use of IP was to strengthen and intensify ties between the university and industry. The resulting acceleration of technology transfer (TT) was said to improve national economic performance.
A comparative survey yields some anomalies:
- the share of national R&D performance accounted for by US and Japanese universities [in those countries] was much lower than that in Italay, the Netherlands, and Canada, and well below the EU average.
- the US government also performs a lower share of non-academic research, and industry a higher share, than is the case in other countries.
- in 1997, 82.5 percent of researchers were employed by industry in the US. The EU average was under 50%.
I concluded that this means the BD route from U to industry is not so important as a share of overall R&D in the country most associated with that route.
Mowery and Sampat say that Us are in fact more tied to industry than is the case elsewhere, though why they say this isn’t clear given their lower share of R&D, etc.
One clear effect of Mode 2 or the Triple Helix or intensified U-I relations however labeled seems to be increased coauthorship (sec 8.3.2). The increases here in the 1980s are remarkable, and they seem to have continued (a doubling of coauthorship by academic researchers in the US bewteen 1981 and 1994 for example). This is the universally cited nano-indicator as well, and yet it long redates nano proper.
their summary of the impact of university research on industry find the following:
- different industries have very different relationships to academic research and to IP. The main BD-style case of TT appears to be the biomedical sector.
- in “other technological and industrial fields, universities occasionally contributed relevant “inventions,” but most commercially significant inventions came from nonacademic research. The incremental advances that were the primary focus of the R&D activities of firms in these sectors were almost exclusively the domain of industrial research, design, problem-solving, and development.” (sec 8.4).
government labs or universities were used more frequently by U.S. industrial firms (on average, in 29.3% of industrial R&D projects) than prototypes emerging from these external sources of research (used in an average of 8.3% of industrial R&D projects). A similar portrait of the relative importance of different outputs of university and public-laboratory research emerges from the responses to questions about the importance to industrial R and D of various information channels (Table 2). Although pharmaceuticals once again is unusual in its assignment of considerable importance to patents and license agreements involving universities and public laboratories, respondents from this industry still rated research publications and conferences as a more important source of information. For most industries, patents and licenses involving inventions from university or public laboratories were reported to be of very little importance, compared with publications, conferences, informal interaction with university researchers, and consulting.”
- “Finally, the channels rated by industrial R&D managers as most important in this complex interaction between academic and industrial innovation rarely include patents and licenses.”
(1) dovetail with Edgerton’s “Shock of the Old”: most “innovation” is actually low-tech;
(2) confirm that university research is most valuable to further research rather than to product development and to economic impacts;
(3) suggest that BD’s focus on IP and linear transfer of academic research has been misplaced.
This last finding has been bourne out by Mowery et al in Ivory Tower and Industrial Innovation (2004) – see my review elsewhere on this site. This would be huge news, if policymakers and PIs really thought about it.
The paper ends with consideration of the utility of two major commercialization strategies.
A. regional economic clusters via university-centered science parks, etc.
- “Using data on U.S. science parks, Felsenstein (1994) finds no evidence that firms located on university-based science parks are more innovative than other local firms, and Wallsten (2001) finds that science parks have a negative effect on regional economic development and rates of innovation.”
- Recent work by Sturgeon (2000) argues that Silicon Valley’s history as a center for new-firm formation and innovation dates back to the early decades of the 20th century, suggesting that much of the region’s innovative “culture” developed over a much longer period of time and predates the ascent to global research eminence of Stanford University. Similarly, the North Carolina “Research Triangle,” which was promoted much more aggressively by the state government, was established in the late 1950s and became a center for new-firm formation and innovation only in the late 1980s.
- Still other work on the development of Silicon Valley by Leslie (1993, 2000) and Saxenian (1988) emphasizes the massive increase in federal defense spending after 1945 as a catalyst for the formation of new high-technology firms in the region. In this view, the presence of leading research universities may have been necessary, but was by no means sufficient, to create Silicon Valley during the 1950s and 1960s. Saxenian in particular emphasizes the very different structure of British defense procurement policies in explaining the lack of similar dynamism in the Cambridge region.
The implication is that “little evidence supports the argument that the presence of universities somehow “causes” the development of regional high-technology agglomerations. And even less evidence supports the argument that the regional or innovation policies of governments are effective in creating these agglomerations.”
I would stress that what seems more important is the “massive increase in federal defense spending,” and emphasize the word massive – huge floods of money making new scales of work and radically experiemental ideas possible to pursue – a certain “throwing money away” exuberence that may help create new things and was certainly part of 1990s Internet culture.
B. Patenting pubicly funded academic research: “Rather than emphasizing public funding and relatively liberal disclosure and dissemination, the Bayh-Dole Act assumes that restrictions on dissemination of the results of many R&D projects will enhance economic efficiency by supporting their commercialization. In many respects, the Bayh-Dole Act is the ultimate expression of faith in the “linear model” of innovation—if basic research results can be purchased by would-be developers, commercial innovation will be accelerated.”
- the patent boom seems closely tied to the boom in biomedical university patents (up 295% vs. up 90% for patents overall in the 1970s).
- OECD and other affirmers of the impact of patenting do not “cite any evidence in support of their claims beyond the clear growth in patenting and licensing by universities. Nor does evidence of increased patenting and licensing by universities by itself indicate that university research discoveries are being transferred to industry more efficiently or commercialized more rapidly, as Colyvas et al. (2001) and Mowery et al. (2001) point out.”
Our CNS research group has confirmed this latter point with our study of Quantum Dot patents, in which quantitative increases in patents have translated into only minor commercial developments. Again the BD implementation of the linear model of innovation appears much less effective and important than is usually assumed.
I’ll mention 3 final epistemological issues that frame the piece, and which I think are broadly relevant to innovation studies:
I. the evidence of studies such as those cited here appear to have little or no impact on police. 5 years after this piece came out, I think it is still true. Work like Gerald Barnett’s on non-IP centered (or altered-IP) mechanisms of early-stage collaboraton is still a hard sell.
II. the evidence of BD and other innovation system impacts is incomplete, and also faulty in its “curent emphasis on the countable” as primary and conclusive evidence.
II. the “emulation” of the US model elsewhere is based on “selective ‘borrowing'” in very different contexts, and is likely to be counterproductive.
One good topic for the workshop would be how 2005-2010 research has or has not modified the situation lucidly anatomized here, particularly around NST.