Academic Inventors, Technological Profiles and Patent Value: An Analysis of Academic Patents Owned by Swedish-Based Firms

Abstract

This paper analyses the relationship between academic inventors and firms, focusing on the relation between academic inventors, the technological profiles of firms and patent value. In particular, this paper focuses on the value of academic patents as compared to non-academic patents, owned by large firms based in Sweden. One finding is that academic patents have a short-term disadvantage, which disappears in the long term. Our results also indicate that controlling for whether the patent belongs to a core or non-core technology relative to the firm's technological profile neutralizes the premium of non-academic patents. In other words, patents belonging to firms' core technologies have significantly higher value, regardless of whether they are academic or non-academic patents. The above results indicate that the technological profile of firms is an important characteristic to analyse, when examining the value of academic patents and the specific role that academics play in industrial invention.

Keywords:

• academic patenting
• patent value
• technological profiles

Keywords:

• O31
• O34
• O39

Acknowledgements

The authors recognize support from the European Science Foundation ESF APE-INV project; University of Gothenburg; and the companies supporting the professorship in Innovation and Industrial Management as well as the earlier EU project KEINS. We acknowledge useful comments from participants at the Schumpeter 2012 conference and the 2012 ESF Leuven workshop. We also acknowledge Yitchak Haberfield, who has provided valuable comments on the econometrics.

Notes

¹ “Academic patent” denotes a patent with one or several inventors affiliated to a university, regardless of assignee. “University patent” denotes patents owned by a university.

² Patent value has in previous studies mostly been measured as the number of forward citations, but in a few instances in monetary terms (e.g. Crespi et al., 2010).

³ Value should here therefore be considered in terms of the technological importance of the patent.

⁴ The teacher's exemption (professor's privilege) means that employees of universities and public research organizations retain intellectual property (IP) rights over the results of their research as individual inventors. The individual can choose to transfer their IP to the university but the university has no claim to IP rights.

⁵ Short-term citations have been shown to be related to the importance and economic value of patents (Lanjouw and Schankerman, 2004), while it has been suggested that later citations indicate the degree to which the patent is “science-based” (Sampat et al., 2003).

⁶ For a detailed account of the KEINS database, constructed in 2004, see Lissoni et al. (2006). This database was expanded by the authors, during the APE-INV project, by collecting new lists of Swedish university researchers in 2011 and matching these to EPO applications. The methodology used largely corresponds to the one employed for constructing the KEINS database. Due to the methodology, we have, in the present sample, not been able to identify those patents invented by academics that retired or changed profession before 2004.

⁷ In order to be able to distinguish between the technological profiles of the firms, we sample firms that consistently and continuously patent, which means that we sample organizations with substantial patent portfolios.

⁸ Webb et al. (2005) present an earlier version of the database.

⁹ We do not exclude self-citations, neither at the inventor level nor at the owner level, i.e., those citations where the citing and cited patents have at least one inventor or owner in common. Excluding self-citations is not possible with our data. We have manually checked the number of self-citations of a small sample of patents. This test indicated that the number of self-citations is not high and that including them does not affect our results in any significant manner.

¹⁰ There is some arbitrariness when drawing the line between short-term and long-term citations, with both three years and five years being used in the literature (Czarnitzki et al., 2012; Sterzi, 2013). Sterzi (2013) finds that the number of citations up to three years after the priority year in corporate patents is on average about 44 per cent higher than that in university patents. Since our study is limited to corporate patents, we use the three-year threshold.

¹¹ We have conducted the same analysis using different levels of patent classifications (IPC3, IPC2, OST7/30), with consistent results.

¹² Although we follow previous studies, there is some arbitrariness in regard to defining the threshold limits. We have tested our models with different threshold values, within reasonable ranges, with overall consistent results.

¹³ The distribution of the technological profiles in our data results to negligible numbers in the “background” and “niche” technological profiles. We therefore limited our analysis to core and non-core technologies including only the dummy “Core” in our models.

¹⁴ See Tables S2 and S3 in the Supplementary material for further descriptive statistics and correlation matrix.

¹⁵ Since we do not control for age, the descriptive statistics of FPC > 3 and FPC require evaluation together with the upcoming econometric models where we do control for age.

¹⁶ Figure 2 is not comparable with the results in Table 3, since in the figure we calculate the average citations per patent by age for all patents, while we in Table 3 calculate the average of the total number of citations received by all patents by June 2012.

¹⁷ This to some extent indicates that the value of academic patents, on average, has declined after the mid-1990s, as have been suggested by previous studies (e.g. Czarnitzki et al., 2011).

¹⁸ The three dependent variables used in our models are count variables.

¹⁹ We have in preliminary models tested the interaction term between the two main determinants, and it appears insignificant. Due to the relatively small amount of academic patents in the sample, the interaction term (of two dummy variables) is highly correlated (0.62) with the academic inventor variable. Thus, adding the interaction term does not add any further effect than the highly significant core variable already adds. We did not include the interaction term in our models because of the multicollinearity problems it introduces and the small value added in the case of our sample.

²⁰ The difference in coefficients between Models 1 and 3 is significant at the 1 per cent level, according to a z-test.

²¹ The difference in coefficients between Models 2 and 4 is significant at the 1 per cent level, according to a z-test.