Government sponsorship and nature of patenting activity of US universities and corporations

Abstract

This paper studies the relationship between government sponsorship and nature of innovation produced by US universities and corporations. Using detailed patent data information and, in particular, from the patent document wrapper, where the applicant is obliged to disclose any federal support, we examine whether (i) federally funded patented innovations are more basic than their non-federally funded peers, and (ii) federally funded corporate and university patented innovations are very different from their existed research agenda. Our results strongly support that federally funded corporate patents are more basic in nature, while the evidence for universities is less nuanced. Also less pronounced and conclusive are the findings about university patented inventions and their ties to university's own research agenda. Results, however, may vary depending on university (corporation) size. While the federal government finances high-risk basic projects, it appears that some firms do not incorporate them in their overall research portfolio.

Keywords:

• government funding
• patents
• basicness
• appropriability

JEL Classification::

• L53
• O31
• O32
• O38

Acknowledgements

We gratefully acknowledge helpful comments received at innovation seminars and workshops. We particularly benefited from valuable comments by David Mowery, Andreas Panagopoulos, Jeffrey Perloff, Brian Wright, and Zhen Lei in earlier versions of the manuscript. We thank Antonia Kosmidi and Konstantinos Vasileiou for research assistance. Kyriakos Drivas gratefully acknowledges financial support from the National Strategic Reference Framework (No: SH1_4083). The usual disclaimer applies.

Notes

1. See theoretical contributions of Romer (1990), Grossman and Helpman (1991), Aghion and Howitt (1998), Jones (2005), and Aghion and Howitt (2005). The relationship between innovation and growth has been also established empirically in many ways and at many levels (e.g. firm, industry, and country levels). See empirical contributions of Coe, Helpman, and Hoffmaister (1997), Keller (2002), Scarpetta and Tressel (2002), Griffith, Redding, and van Reenen (2004), and Cameron, Proudman, and Redding (2005).

2. Government R&D and innovation policies are perceived as crucial elements of many success stories in the private sector around the world. Of particular interest is the Israeli experience because of its boomed high-tech sector and considerable subsidization of R&D activity there (Lach 2002; Trajtenberg 2002).

3. One leading example of a long-term funding project, initially funded from the US Defense Advanced Research Projects Agency in the early 1960s, and then later by the US National Science Foundation, was the development of the Internet that has massively improved the lives of many people in the USA and around the world. Internet-related technologies and businesses have also developed as the result of federal support, including Google. Furthermore, government support for research in medical science has led to the creation and expansion of the biotechnology industry with particularly important benefits given its impact on the expected length and quality of life.

4. Universities also fund their research by non-federal funds (i.e. universities’ and colleges’ own institutional funds, state and local government funds, industry funds, non-profits funds, and other organizations’ funds). Nevertheless, the federal government remains the largest funder of their research. For example, the federal government provided 59% of academic spending on science and engineering R&D in 2009.

5. Henderson, Jaffe, and Trajtenber (1998) document a sharp increase in university patenting accompanied by a decrease in the relative ‘importance’, in terms of path-breaking, of university innovation over the later part of their 1965–1992 sample.

6. For example, THJ find that university patents are more basic in nature compared to the corporate patents and that corporate patents, in general, build significantly more on own patents, and, therefore, have higher degree of appropriability, than university patents; however, they do not distinguish between federally versus non-federally funded patents and do not account for different university or corporation sizes.

7. The ‘exploitation’ of the information provided on the patent document wrapper was actually motivated by the work of Pressman et al. (2006) who used information of the patent document to identify which DNA patents had disclosed support from NIH to examine whether NIH licensing guidelines are violated.

8. For instance, many of the big oil companies are financed, in one fashion or the other, by the federal government to conduct research on renewable resources, for example, biofuels (Wahburn 2010). While this area is somewhat unknown to those firms, it could well become very familiar to the firms in the future.

9. For more details, see https://sites.google.com/site/patentdataproject/Home.

10. Alcacer and Gittelman (2006) show that two-thirds of the prior art is inserted by the examiner.

11. Forward citations are citations to a given patent from patents that are ‘forward in time’ from this patent. In contrast, backward citations to a given patent are patents that are ‘backward in time’ from this patent. Therefore, forward citations capture the prior effects of an innovation, while backward citations capture the posterior effect of an innovation. We use backward and forward citations to assess the nature of patents.

12. For a discussion about the notion of basicness and appropriability of research outcome, see Trajtenberg, Henderson, and Jaffe (1992).

13. For instance, Lowe (2002) has used this metric to examine whether more basic university patented inventions are more likely to be licensed to start-up firms than established firms

14. For both metrics, ImportB and ImportF, we choose λ=0.5 as in THJ. In Table A1, we offer results also for values and .

15. We consider 419 US Classifications in our sample.

16. Indeed, two recent papers, one by Jacob and Lefgren (2011), who use instrumental variables to examine the impact of NIH on productivity of academic scientists, and one by Azoulay, Zivin, and Manso (2011), who employ propensity score matching to compare the effect of an NIH and Howard Hughes Medical Institute research grant on scientific productivity, have employed such techniques since their data disaggregation allows them to do so. Although there is some disaggregated information for universities, there is no much, to our knowledge, about corporations.

17. The database is available at http://sites.google.com/site/patentdataproject.

18. There are additionally 1716 patents that are awarded to at least one US-located firm and US-located university and 383 (22.3%) of them have disclosed federal support. We have examined this sub-sample, but due to the relatively small size, it is difficult to draw meaningful inferences. Results are available upon request.

19. Our sample includes 336 US universities, the vast majority of the US universities that produce patents, and 35,656 firms.

20. The dataset is available at http://dvn.iq.harvard.edu/dvn/dv/patent.

21. For example, for patent 6,015,106, the statement is: ‘This invention was made with Government support under Agreement No. NMA202-97-9-1050 awarded by the National Imagery and Mapping Agency. The Government has certain rights in the invention’. In another example, for patent 6,030,819: ‘This invention was made with Government support under Government Contract No. 70NANB5H1135, awarded by the National Institute of Science and Technology. The Government has certain rights in this invention’. Pressman et al. (2006) use this information to identify which DNA patents had disclosed support from the NIH and examine whether NIH licensing guidelines are violated.

22. Even though NIH has received three petitions to exercise ‘march-in’ rights (BayhDole25 2006), they have never been exercised to this date.

23. Out of 4345 patents assigned to small universities, 1350 (31%) disclose government sponsorship; out of 4439 patents assigned to medium universities, 1835 (41%) are federally funded; and finally, 2440 (57%) patents out of 4286 that are assigned to large universities disclose federal support.

24. Out of 94,699 patents assigned to small firms, 1217 (1.3%) disclose government sponsorship; out of the 95,720 patents assigned to medium firms, 1969 (2.1%) are federally funded; and finally, out of 1136 patents that are assigned to large firms, 95,997 (1.2%) disclose federal support.

25. The division rule we applied is the following: large-sized entity > 66th percentile of the university (corporate) patents distribution; small-sized entity < 33rd percentile of the university (corporate) patents distribution. For robustness purposes, we have also considered a different cut-off point: large-sized entity > 75th percentile of the university (corporate) patents distribution; 75th percentile > medium-large-sized entity > 50th; 50th percentile entity > medium-small-sized entity > 25th percentile; and small-sized entity < 25th percentile of the university (corporate) patents distribution. See Tables A2 and A3 for university and corporation results, respectively.

26. This patent measure is not available in THJ; the coefficient we display here agrees with intuition.

27. Table A1 reports results for different values of λ. We report results for and . As the table displays, results are qualitatively similar. Since results do not alter, for the rest of the paper, we only display results for λ=0.5.

28. For robustness purposes, we have also considered different cut-off points. See Table A2. Results, overall, do not change.

29. For robustness purposes, we have also considered different cut-off points (Table A3). Results, overall, do not change.

30. For an elaboration, see Ai and Norton (2003).

31. Calculations are available upon request.

32. For robustness purposes, we have also considered different cut-off points (Table A2). Results, overall, do not change.

33. For robustness purposes, we have also considered different cut-off points (Table A3). Results, overall, do not change.