The impact of the abolishment of the professor’s privilege on European university-owned patents

ABSTRACT

Intellectual property regimes governing university inventions were quite diverse in Europe at the end of the 1990s. Several European countries maintained the so-called professor’s privilege, an exception to employment law whereby university researchers were allowed to retain the ownership of academic inventions. The 2000s were characterised by convergence towards a more homogeneous system, in which university administrations took control of IP management. We investigate the impact of the reform and we observe a decline in the technological importance and the value of the patents owned and managed by universities in the countries abolishing the professor’s privilege. On the contrary, by differentiating the academic patents by type of ownership, we find that the technological importance of academic patents owned by companies has instead increased. Our study produces some new results that may alert policymakers to the possible unintended consequences of the university ownership model.

KEYWORDS:

• University-owned patents
• academic-invented patents
• patent quality
• patent management
• patent value
• technology transfer

Acknowledgments

We gratefully acknowledge funding from the French Agence Nationale de la Recherche (UTTO Project ANR-15-CE26-0005 and NPEIE Project ANR-17-CE26-0014-01), the Spanish Ministry of Economy and Competitiveness (CSO2016-79045-C2-1-R), the Regional Government of Madrid (PRODECON-CM, S2015/HUM-3491) and the CNRS-CSIC 2018 IRP ALLIES - Associated Laboratory on Linkages between Innovation and Environmental Sustainability. We wish to thank Nicolas Carayol, Elodie Carpentier, Chirantan Chatterjee, Alberto Galasso, Marco Giarratana, Francesco Lissoni, Gerard Llobet, Manuel Mira-Goudinho, Michele Pezzoni and seminar participants at the 2017 EPIP Conference, Bordeaux, at the 2017 PRODECON-CM Conference on Innovation and Industrial Organization, Madrid, at the 2019 Collegio Carlo Alberto Seminar in Turin, at the 2019 Catedra per al Foment de la Innovació Empresarial Seminar at Universitat Rovira I Virgili in Reus and at the 2019 University of Insubria, Economics Department Seminar, in Varese for helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ According to the linear model of technology transfer, the process begins with a discovery by the university researcher (laboratory) and follows a linear path from disclosure to the university administration (possibly to the TTO) overseeing the patenting process and, finally, to licences to an existing firm for further development and commercialisation (see Bradley, Hayter, and Link 2013 for reviews of the literature on technology transfer).

² The influence of the Act has also extended to other parts of the world, in addition to Europe, both in developed countries such as Japan (Takenaka 2005) and in emerging countries such as Malaysia, South Africa or the Philippines (Zuniga 2011).

³ These march-in rights have not been exercised at the time of writing.

⁴ Damsgaard and Thursby (2013), Greenbaum and Scott (2010), Grimaldi et al. (2011), Kenney and Patton (2011), Mowery and Sampat (2005), Schacht (2012), Valentin and Jensen (2007), Von Proff, Buenstorf, and Hummel (2012).

⁵ Two previous descriptive studies, using data from Norway, find less drastic decreases. Iversen, Gulbrandsen, and Klitkou (2007) distinguish between patents with inventors affiliated to Norwegian universities and patents with inventors affiliated to public research institutes and filed between 1998 and 2003. They observe a decrease in 2003, but argue that this is temporary and attribute it to a lack of experience and uncertainty in face of the change. They further note that ‘the conception of “professor privilege” patents as predominantly single-inventor patents held by the researcher himself tends to be inaccurate in most sectors. Individual public sector research inventor patents are in fact relatively limited.’ Using a larger sample, Spilling et al. (2015) show that the share of patent filings having inventors affiliated to the Norwegian public sector fell slightly (about 1%) between 1999–2003 and 2004–2008, but they warn against attributing it to the policy change, as the overall economic context changed greatly during this period in Norway.

⁶ Valentin and Jensen (2007) use data from the Scanbit Database from the Research Centre on Biotech Business at Copenhagen Business School, which includes patent information and other indicators of pharmaceutical discovery by firms dedicated to biotechnology in Denmark, Sweden and Norway.

⁷ In descriptive analyses for Germany, Schmoch (2007) observes no evidence of a change in the numbers of academic- invented patents before and after the abolishment of the privilege, while Dornbusch et al. (2012) point to a different trend in patent filings made at EPO and the German Patent Office by German academics. However, EPO filings, which are likely to be considered of a higher expected value by applicants, increased after the change, whereas filings to the German patent office decreased.

⁸ As we will see later, we exclude Finland from our econometric analysis because it abolished the professor’s privilege in 2007, some years later than the other IP reform countries we consider.

⁹ We may also expect changes to the IP regime to have little or no effect on researchers’ inventive efforts. This would be the case if university researchers abide by Mertonian norms of science rather than financial rewards (Stephan 1996; Evans and Leighton 1989; Hamilton 2000; Shane, Locke, and Collins 2003).

¹⁰ From this point of view, the abolition of the professor’s privilege may also induce inventors to begin patenting.

¹¹ We do not consider in the analysis patents filed after 2008 in order to have enough years of observation for collecting data on forward citations that we use to proxy the technological importance of the inventions.

¹² We use the terms ‘university-owned patents’ or ‘university patents’ to refer to patent applications originally owned by universities at the time of the priority filing. Patents are thus classified by first ownership. Alternatively, patents can be classified by the affiliation of inventors, and distinguish between patents invented by academics but owned by companies and patents invented by academics and owned by universities, the latter are the university-owned patents we focus for the most part of the present paper, but for completeness, we also include an analysis of academic-invented patents, comparing company-owned and university-owned in Section 7 later.

¹³ As included in Table tls906_person of PATSTAT April 2016.

¹⁴ University-company co-applications (categories #5 and #6) are excluded from the analysis as they represent only a small share of patents with at least one university as applicant and the origin and management of the inventions is unclear. However, they represent an increasing share of patents in all countries and the results do not change when co-applications are include.

¹⁵ However, the inclusion of early citations does not change the main results of the paper.

¹⁶ As an alternative measure of the technological importance of the invention we use Patent Originality. This variable is taken from the OECD Patent Quality Dataset (Squicciarini, Dernis, and Criscuolo 2013, database version 2017) and refers ‘to the breadth of the technology fields on which a patent relies’ (Squicciarini, Dernis, and Criscuolo 2013). This indicator, proposed by Henderson, Jaffe, and Trajtenberg (1998) and refined by Hall, Jaffe, and Trajtenberg (2005), is based on the dispersion of the technological classifications of the backward citations and varies between zero and one, with higher values for more original patents. We acknowledge that the concept of patent originality is closer to technological diversity than to technological importance (Jaffe and De Rassenfosse 2017). Our data do however corroborate this intuition; the correlation between the two indicators is only 10%. Consequently, the results of using patent originality as the dependent variable (see Table A2) are quite different from those shown in Table 7. Firstly, university patents do not differ from company patents in terms of originality. The dummy University is positive but not significantly different from zero in almost all the models. Secondly, we observe no difference between university patents in countries abolishing the professor’s privilege and the control group, neither before nor after the reform.

¹⁷ The dummy POST takes the value one in year 2002 and following years for all the countries in the data (treated and controls), with the exception of Denmark (for which the ‘post’ variable identifies the IP regime change in 2000) and Norway (2003).

¹⁸ Given that there are 35 technology fields, we should have 35*18*11 = 6930 observations, but we miss data points because not all countries, fields and years have patents granted at EPO.

¹⁹ Results available upon request.

²⁰ A better method to identify academic patents relies on matching inventors and academic staff lists, but that is fraught with problems and requires an enormous effort beyond the scope of this paper (e.g. difficulty in obtaining staff lists, problems of homonymy, imperfect disambiguation strategies, lack of large scale golden standards to estimate precision and recall rates and overall reliability of resulting datasets). To our knowledge, there have only been two important paneuropean collective efforts in the past to gather data on academic patenting based on matching academic staff lists. The first one, excluding Germany, is the KEINS project in the early 2000s (Lissoni et al. 2008). The second one is the APE-INV project in the late 2000s (Lissoni 2012), to which the authors of this paper have made significant contributions, with studies of academic patenting in the UK, Italy and Spain. German data is available in the APE-INV public datasets but the time range available is limited to 2006–2007 and thus not useful for pour purposes (Schoen, Heinisch, and Buenstorf 2014). Austria is also included, but for a different time period 1997–2009 (Backs, Günther, and Stummer 2019). With the aim to keep a sample as homogenous as possible in the data sampling and identification strategy, we use the ‘Prof.Dr.’ title technique for both Germany and Austria for this new section whose main purpose is to provide some robustness tests.

²¹ Another limitation put forward by von Proff et al (2012) is that top-level R&D employees of R&D intensive firms may hold honorary professorships at universities, but that is a limitation that could also affect other methods to identify academic patents, such as those based on matching names of university professors to names of academic inventors.

²² We thank Nicolas Carayol and Elodie Carpentier for the data.

²³ We also run models by adding as further controls corporate and academic (university- and company-owned) French and Spanish patents. Results do not substantially differ and are available upon request from the authors.

²⁴ The STATA module psmatch2 has been used for matching. We adopt the nearest-neighbour algorithm (Rosenbaum and Rubin 1983) and we match each university patent to a company patent.

²⁵ Three university patents have not been matched to any company patent.

²⁶ After matching (as shown in Table A4), the averages of invention quality (IQ, long-term citations) are not statistically different in the group of university patents and company patents, confirming that the balance of covariates has been successfully achieved; prior to PSM, universities receive 4.62 (long-term) citations and company patents only 3.05. After matching, the difference between the two groups disappears (t-test of 0.73).