Academic Inventions Outside the University: Investigating Patent Ownership in the UK

Abstract

This paper investigates the ownership of academic patents for a sample of UK academics and challenges the existing definition of the university invention ownership model. The first descriptive results show that 50 per cent of patents are owned by industry; however, 37 per cent of these firm-assigned patents are in fact owned by university spin-offs. We investigate how university policy and funding acquisition impacts industry versus university ownership, and find that funding from large firms predicts involvement in patenting and, to a lesser extent, firm ownership. University ownership of academic patents is more likely the higher the amount of funding coming from SMEs, and at universities that outsource the filing of patents. Spin-off patents occupy an intermediate position showing strong similarities to both firm and university patents.

Keywords:

• academic patenting
• university–industry collaboration
• sponsored research
• academic entrepreneurship
• patent ownership

Keywords:

• I23
• 031
• 034

Acknowledgements

The author would like to thank Aldo Geuna, three anonymous referees and seminar participants at APE-INV, Leuven, APIC, Seoul, and MBIR, Manchester, for helpful suggestions and comments. This paper contributes to the FP7 Project SISOB (#266588) and the Collegio Carlo Alberto Project “Researcher Mobility and Scientific Performance”. Sponsorship within the ESF Activity “Academic Patenting in Europe” is also gratefully acknowledged.

Notes

¹ Kenney and Patton (2009) question the efficiency of this university ownership model that has made universities to “revenue maximizers, rather than facilitators of technology dissemination for the good of the entire society” (p. 1407).

² Though US universities are more successful in licensing, UK universities create relatively more spin-off companies (HEFCE, 2011).

³ Universities follow different strategies in managing IP. Some offices may only patent the most promising inventions (Meyer and Tang, 2007), while others may prefer to invest in spin-offs or choose not to enforce their rights (Markman et al., 2008).

⁴ For a list of universities see Supplement S1.

⁵ For a detailed description of the original data, see Banal-Estanol et al. (2010). See Lawson (2013) for an analysis of sample distribution differences between the original data and a sub-sample of researchers at 10 UK universities, which partially overlaps with the data used in this paper.

⁶ The full process of extracting and cleaning the data is described in Meissner (2011).

⁷ We do not have information on workplaces for all academics during the period 1998–2000 and therefore cannot make inferences on whether patents are academic patents or not.

⁸ Ownership can change across time and across patent offices. In the case of regional sister patents, ownership can differ from that of the original application. This is ignored, as this extension happened at a later date. Even if we consider ownership on all sister patents across time, only 15 patents are ‘co-assigned’ to different types of agents. This low number is partly due to the short time window with most applications not having entered regional stages at the time of data collection.

⁹ FAME is a company information register for the UK.

¹⁰ We identified university spin-offs but they were only responsible for 20 grants during the observation period.

¹¹ Other measures are not available or do not differ across institutions. For example, several papers have found that revenue incentives paid by universities increase disclosure activity of academic staff (Jensen and Thursby, 2001; Lach and Schankerman, 2008). However, revenue shares in the UK typically follow a structured system with lower returns for higher licensing amounts and are similar across universities. Therefore, we do not include them into the regression. Also, Markman et al. (2008) find no significant effect of licensing shares on university ownership when including other university characteristics.

¹² Information was taken from personal websites or Index to Theses, a listing of theses accepted for higher degrees in the UK and Ireland.

¹³ Most patents in our data are still being examined and have not as yet entered regional phase (in case of Patent Cooperation Treaty patents), which further makes the use of patent-based measures problematic. Only 44 patents had been granted when the data was collected in 2012, either at the EPO (14) or at a regional office, and only 2 EPO patents had received any citations.

¹⁴ In fact we are looking at 244 patents; 23 patents have more than one inventor in the data set. We consider these multiple counts of the same patent as explanatory variables differ across researchers.

¹⁵ Ownership can lie with the government or solely with an individual but these account for just 23 observations and are not included as separate categories.

¹⁶ We use the user-written command cmp in Stata to estimate the bivariate probit models with selection (Roodman, 2009)

¹⁷ A multivariate probit with selection was not possible as the model would not converge. As an alternative one could perform separate probit estimations and indeed the results are confirmed. The descriptive statistics for the reduced number of observations can be found in S5.