- Split View
-
Views
-
Cite
Cite
Valerio Sterzi, Michele Pezzoni, Francesco Lissoni, Patent management by universities: evidence from Italian academic inventions, Industrial and Corporate Change, Volume 28, Issue 2, April 2019, Pages 309–330, https://doi.org/10.1093/icc/dty070
- Share Icon Share
Abstract
Over the past 20 years, European universities have increased their propensity to retain title of their faculty’s inventions, but evidence on the value of such patents is at best mixed. Based on a longitudinal sample of Italian academic patents (patents over faculty’s inventions), assigned either to universities or firms, we find that the lower value of university-owned patents, versus firm-owned ones, is owing to lower Technological Importance of the inventions and less effective Exploitation of the related patents. Lack of experience in managing patented inventions explains our results for Technological Importance, but not for Exploitation. Both are unrelated to the presence of a technology transfer office. Our study suggests caution in pushing universities to expand their patent portfolios and in using university-owned patents as indicators of technology-transfer activities.
1. Introduction
Over the past 20 years, European universities have been constantly pushed to reclaim intellectual property (IP) either from their employees or from industry partners. Two good illustrations of this trend are the “Code of Practice” approved by the European Commission in April 2008, which explicitly recommends universities to create “coherent portfolios of intellectual property” (European Commission, 2008; Arundel et al., 2013), and the abolition of laws prescribing the “professor’s privilege”, which, in all German-speaking and Scandinavian countries, exempted academics from any disclosure duty toward their university of affiliation (Czarnitzki et al., 2015; Hvide and Jones, 2018). More importantly, autonomy-granting reforms throughout the continent have provided universities with both the legal means and the economic incentives to pursue more aggressive IP appropriation policies (Estermann and Nokkala, 2009; Lissoni et al., 2013).
Universities’ direct involvement in IP management is usually justified by the necessity to “better convert knowledge into socio-economic benefits” (in the words of the European Commission’s press release IP/08/55). Yet, little or no evidence suggests that university administrations, among all actors involved in their faculty’s inventive process, are best placed to decide which academic inventions are worth patenting and/or to manage the resulting patent portfolio.
Recent European cross-country evidence, based on citation metrics, shows that university-owned patents over academic inventions are most often of lower value than firm-owned ones (Lissoni and Montobbio, 2015). Evidence from country-specific studies is at best mixed (Czarnitzki et al., 2012; Callaert et al., 2013; Sterzi, 2013). Recent US-based evidence also suggests that corporate-sponsored academic inventions, whose IP rights most often remain with industry, have higher value and impact than those funded only with federal grants (Wright et al., 2014). Still, this literature does not distinguish between two potential sources for the lower value of university-owned patents, namely, a lower commercial potential of the patented inventions and/or a lack of effectiveness when it comes to commercializing them (a distinction we adopt from Bessen, 2008). The former may be owing to the limited relevance or originality (Technological Importance) of the academic inventions that end up in the universities’ portfolio, relative to those assigned to firms. The latter refers to the activities undertaken to exploit the assigned patents (Exploitation), which universities may find more difficult to undertake, owing to lack of managerial skills, means, or incentives, relative to firms. Although logically distinct and originating from activities that take place at different points in time in the life of an academic patent, the consequences of Technological Importance and Exploitation for the value of patents are often measured with the same indicators (the most common being forward citations) or with highly correlated indicators (e.g., citations and the number of patent renewals). We contribute to the literature with an empirical exercise that investigates both causes and their relative importance at the same time, and how this may depend on the patenting experience accumulated by universities over the past few decades.
We focus on the case of Italy, which is both representative of European trends and interesting per se, as witnessed by the considerable scholarly attention it has attracted (surveys by: Geuna and Rossi, 2011; Lissoni, 2012; Kochenkova et al., 2016). Based on European Patent Office (EPO) data for 1995–2001, Italian patents over academic inventions (academic patents) stood at over 1100, as many as the Dutch ones, significantly more than the Swedish, and two thirds of the French (Lissoni and Montobbio, 2015). Data collected with the same methodology for the following years show a remarkable growth in the university-owned share (Lissoni et al., 2013), despite the extemporaneous introduction of the professor’s privilege in 2001, subsequently reformed in 2005 (and fiercely opposed by university administrations; see Baldini et al., 2010).
We find that Italian universities both own inventions of lower Technological Importance and extract less value from the related patents. We also find that their relative lack of experience in managing patents, due to a late start in handling IP, may explain our results in the first case, but not in the second one. The presence of a technology transfer office (TTO) in the university at the time of the patent filing never makes a difference.
Policy-wise, our findings cast doubts on any measure resulting in more pressure on universities to increase their patent portfolios, either by disputing to firms the ownership of academic patents or by filing applications on academic inventions of no interest to firms.
In the remainder of the article we proceed as follows. In Section 2, we review the existing literature on academic patenting, provide some historical information on the phenomenon in Italy, and put forward our distinction between Technological Importance and Exploitation. In Section 3, we present our data, variables, and estimation strategy. Section 4 reports and discusses the results. Section 5 concludes.
2. Background literature and theoretical framework
We define as “academic” all the inventions that originate from research performed by one or more university researchers as part of their job, whether or not in collaboration with industry, and regardless of the ownership of the resulting patents. As discussed by Lissoni et al. (2009), academic inventions may be patented by the inventors themselves, their universities of affiliation, some private or public sponsors (most often, respectively, business companies or governmental agencies), or a combination thereof. When a patent application is filed, academic inventions generate corresponding “academic patents,” which we can further distinguish between “university-owned,” “firm-owned,” “inventor-owned,” and so forth, depending on the applicants’ identity.1
Our terminology is coherent with recent, European-oriented literature (Geuna and Rossi, 2011; Kochenkova et al., 2016) and compatible with a nonlinear view of the generation of academic inventions, which owes to Nathan Rosenberg’s view of the interaction between science and technology, as well as university and industry (Rosenberg, 1982; Landau and Rosenberg, 1986; Rosenberg and Nelson, 1994). According to such a view, academic inventions do not occur at the end of an imaginary linear path, running from public-funded basic research to one or more proofs of concepts or prototypes (as in Jensen and Thursby, 2001). Rather, they stem from the frequent interaction between several actors, ranging from the inventor herself, her university administration, and the business contacts brought in by one or the other. This is in accordance with Rosenberg’s historical account of several US academic inventions, whether university-owned (as in Colyvas et al., 2002) or firm-owned (such as the first laser patents described by Rosenberg, 2003, based on Townes, 1999). Far from being a distraction from fundamental research, interaction with companies may counter academic researchers’ conservatism opening up new promising fields of research and result in major technological advancements (Evans, 2010; Stokes, 2011; see also evidence by Wright et al., 2014).
When it comes to IP, such interaction originates ownership issues at several points in time, ranging from the moment when the academic inventor receives any private funds (possibly in exchange of IP rights or options on research results yet to come) to that of patent filing (when an industrial partner may be brought in to cover the patent fees, in exchange of IP ownership or licensing). Several inventions discussed by Colyvas et al. (2002), albeit concerning academic patents ultimately retained by US universities, obtained IP protection under the pressure exerted by their industrial partners, who had sustained or followed the inventive process from its early stages and were eager to get licenses.
Although the theoretical roots of our work and most of the European literature on academic patents lie in what we call a nonlinear view of academic inventions, their policy motivations originate in the critical review of patent indicators that informed the debate on technology transfer in the early 2000s. At a time when policymakers lamented the dearth of inventive activity by European universities, as measured by university-owned patents, the diffusion of a new, inventor-based data-mining technique rapidly uncovered a rather different picture.2 Far from being negligible, academic patenting in Europe accounts for around 5% of most countries’ patenting activity, with peaks of over 20% in biotechnology, very close to US figures (1994–2002 data; as reported by Lissoni, 2012). At the time of the earliest studies, European academic patenting was characterized by a low rate of university ownership and a corresponding high rate of firm ownership (60–80% firm-owned vs. 5–25% university-owned for France, Italy, the Netherlands, Sweden, and the UK, as opposed to about 25% vs. 69% in the United States; Lissoni et al., 2009). This pattern stemmed from the European university administrations’ relative lack of interest into IP and lack of means for exploiting it, as well as in the formal or informal control exerted on such IP by the academic inventors themselves. The following studies, however, have revealed an increasing trend toward university ownership, due ultimately to two policy changes: first and foremost, the increased autonomy conceded to many European universities (Arundel et al., 2013), and, second, the abolition of legal provisions going under the name of professor’s privilege.3
Autonomy-increasing policy measures date back to the late 1980s (Estermann and Nokkala, 2009; Christensen, 2011). They have progressively allowed universities to manage their own financial assets and, at the same time, they have shifted public funding from block grants to performance-based and competitive financing. Besides, they have increased the pressure on universities to raise their own funds, also through the commercialization of research results (Geuna and Martin, 2003; Geuna et al., 2015; Jonkers and Zacharewicz, 2016). Under these circumstances, universities have started regarding IP both as a possible source of revenues and as a way to signal their engagement in technology transfer both to potential partners from industry and to public funding agencies. Autonomy has also given universities more control over the activities of their faculty, whose civil servant status, or at least some prerogatives that went with it, considerably weakened (Enders and Musselin, 2008). In Italy, universities were granted full managerial and financial autonomy in between 1989 and 1996, and soon afterwards adopted measures such as opening a TTO, introducing administrative norms aimed at taking control of IP (from now on: IP regulations), and participating to spin-off firms (Baldini et al., 2006).4
As for the professor’s privilege, this was part of IP legislations typical of Scandinavian and German-speaking countries.5 Where in force, it exempted academics from any disclosure duty with respect to their university of affiliation and granted them exclusive IP rights over their inventions. Nowadays, it survives only in Sweden. In 2001, it was unexpectedly introduced in Italy, not as part of any IP or university reform but as an extemporaneous amendment to the Budget Law. Following an outcry of protest by the largest universities, it was watered down soon after, as part of a general reform of IP law.6
Several authors have investigated whether the diffusion of the university-ownership model has gone along with an increase or a decrease of the economic value of academic patents, especially university-owned ones.7 The question is relevant both for assessing the viability of measures that push universities to get involved in IP management and for the general welfare effects that such involvement may generate.
Most of the existing studies measure the value of the academic patents by counting the number of forward citations received. Two levels of analysis are possible. First, one can compare academic and nonacademic patents. Alternatively, one can focus on university-owned versus firm-owned academic patents. In this article, we choose this second level of analysis.
Lissoni and Montobbio (2015) examine the citation rates of academic patents in several European countries and find them to differ greatly across country, conditional on ownership. For instance, academic patents in the Netherlands are more cited than nonacademic ones, irrespective of their ownership, whereas university-owned patents get fewer citations in both France and Italy. For the UK, Sterzi (2013) finds that university-owned academic patents are less cited than the firm-owned ones and that ownership changes (from university to firm) increase the citation rate. For Germany, Czarnitzki et al. (2012) find that firm-owned academic patents receive on average more citations than university-owned ones and that this difference is explained only by citations received in the short term.8 As for the United States, Thursby et al. (2009) provide somehow contrary evidence for a sample of 5811 academic patents, 26% of which are assigned solely to firms and get cited in a lower number of technological classes than those assigned to universities.
Although this empirical literature is rather convergent in showing that university-owned academic patents are less valuable than firm-owned academic patents, at least for Europe, it stops short when it comes to explanations. In particular, it does not distinguish between two potential sources of the difference in value that we conceptualize as follows: the characteristics of the academic inventions, such as its technological relevance and originality, which may affect the commercial potential of the related patent (Technological Importance); and the effectiveness of the activities undertaken to exploit the patent by its assignee, whether a university or a firm, on the basis of its experience, means, and incentives (Exploitation).
Figure 1 shows that the two components of the Patent Value are determined at different points in time, namely, before and after the patent filing.
The Technological Importance is an attribute of the academic invention that emerges during the research activities undertaken by the faculty, possibly in collaboration with industry. The decision to file a patent application and the choice of the assignee (the university, a company, or, increasingly less frequently, the inventor or a public sponsor) also emerge during the research activity and potentially involve different actors. Were we to observe a negative association between the Technological Importance of the patent and university ownership (a positive association with firm ownership), it would be difficult to establish a causal link.
At one end, collaboration with industry may jointly lead to more important inventions and to the assignment of the patent to the partner firm, as when a collaboration contract is signed that includes clauses reserving all IP rights to the firm, at least in first instance, and the latter decide finally to exert them.9 In Italy, this is often the case with collaborations involving not just individual faculty but also their departments. These contracts, including their IP clauses, need to be approved by the university administrations but are generally negotiated by the departments in a rather autonomous way.
At the other end, research may be conducted in the absence of firm contacts, at least initially, with the decision on whether to file a patent application, and to whom to assign it, being taken after obtaining the invention. In this case, which is closer to the linear view of academic inventions, some asymmetric information and moral hazard issues may arise, with academic inventors having an incentive not to disclose to the university administration their most important inventions and look instead for an industrial partner, in exchange for a higher monetary reward or secure research funding (Siegel et al., 2007; Markman et al., 2008; Carayol and Sterzi, 2018).
Several situations stand in between, which are also coherent with university ownership being negatively associated to the Technological Importance of the invention, such as when a collaborating firm judges the invention unimportant and does not exert its IP option but the university decides to file a patent application anyway. Recent evaluation exercises, such as the English Research Assessment Exercise (RAE) or the Italian Valutazione Quadriennale della Ricerca (VQR), may indeed provide public universities with an incentive to file patent applications on inventions of dubious importance, to the extent that university-owned patents enter the evaluation indicators.10
Coming to the Exploitation process, this occurs after the patent application has been filed (see again Figure 1) and its success does not depend exclusively on the importance of the invention but also on the commercialization means, experience, and incentives of the patent owners.
Concerning means, we observe that universities, as nonpracticing entities, have a relatively limited set of commercialization options (Lemley, 2007). These are restricted to licensing or selling, whereas manufacturing firms can also proceed to direct exploitation or cross-licensing with competitors. This lack of options might impact negatively on the value extracted from patents, especially in contexts where the national market for technologies is less developed than in the United States, as it is the case for several European countries, including Italy.
Considering incentives, we observe that, wherever university-owned patents enter the evaluation indicators used by funding agencies, universities may obtain some rewards from having filed them even without engaging in active commercialization efforts, which may detract to the incentives of doing so.
Finally, the universities may simply lack the managerial skills to engage profitably in the markets for technologies and do not have the financial or legal means to compensate by hiring expert consultants.11
The empirical literature we reviewed so far did not consider the Technological Importance of the invention as distinct from the Patent Value created through effective exploitation, for two main reasons. First, the most common indicator we find in the literature, forward citations, is used alternatively to capture the so-called “quality” of the patent (by which most authors mean the characteristics of the invention) or its economic value (which successful licensing or in-house exploitation efforts may increase). We further discuss this point below. Second, we cannot observe the universities’ exploitation skills, which forces us to measure their impact on the Patent Value in an indirect way.
3. Data and estimation strategy
3.1 Data and measurement
We exploit the APE-INV data set collected by Lissoni et al. (2013), to which we refer for full methodological details and in-depth descriptive statistics. The data set contains all patent applications filed at the EPO from 1978 to 2009 and signed by inventors with both an Italian address and a tenured position in an Italian university in 2000, 2005, and 2009. We retain only the applications with priority dates between 1997 and 2009, for a total of 3651 Italian academic patents.12
Data on patents’ first applicants, International Patent Classification (IPC) classification, and citations come from the April 2014 edition of the Worldwide Patent Statistical Database (PatStat). Data on patents’ granted status, citations, and renewal dates come from the OECD Patent Quality Dataset (Squicciarini et al., 2013).13
We classify different types of first applicants as follows:
Firm, when the first applicant is a firm, multiple firms, or a combination of firms and individual inventors (2211 patents, i.e., 60.5% of the sample);
University, when the first applicant is the university of affiliation of one or more of the academic inventors in the patent, either alone or in combination with other universities, public research organizations, or spin-offs with a university participation (700 patents, i.e., 19% of the sample; of which, 491 of exclusive university property);
Faculty, when the first applicant is one or more of the academic inventors listed on the patent, possibly associated with other individuals (340 patents, 9% of the sample);
University&Firm, when the first applicant is the university of affiliation of one or more of the academic inventors in the patent in combination with a firm (93 patents, 2.5% of the sample).
Others, when the first applicant is a foreign university or a public research organization, either Italian or foreign, or a combination thereof (307 patents, 8% of the sample);
Figure 2 illustrates how Italian universities have gradually increased their ownership of academic patents, in response to the institutional and policy changes described in Section 2. Although the number of academic patents owned by universities increases steadily, that of patents owned by faculty and business sector remains steady (the decline in the most recent years is due to right truncation of data).
In our analysis, we consider only the firm-owned and university-owned categories, owing to their prevalence as well as to the attention the literature has paid to both of them. We do not include in the sample the academic patents co-owned by university and firm (less than 2.5% of the observations). This is because in such cases universities are likely to be passive partners, with all management matters in the firm’s hands (the patents are later on taken over by the firm, which either pays for the international extension or buys them outright). However, their inclusion in the data set as part of the university-owned category does not alter our main results.14
When it comes to defining our main variables, we face a number of challenges, starting with Patent Value. This ought to correspond intuitively to the overall revenues generated by the patent. However, getting data for these revenues is either hard or impossible. For the academic patents that assignees exploit through licensing we would need to know the royalty figures, which are mostly unavailable, and for those exploited in-house (which is possible only for firm-owned patents) we would need some figurative calculations of their impact on the assignees’ overall revenues. In addition, we would have to capture both the private and the public value of the patent, to the extent that universities may opt for licensing schemes and fees aimed at maximizing the use of their patents, rather than their own revenues.
Hence, we rely on a proxy used, among others, by Lanjouw et al. (1998), namely, the number of renewals paid for by the patent assignees or, more precisely, the number of years of validity of the focal patent from the filing year (Renewals). To the extent that maintaining the patent protection over time is costly, one can assume that any valuable patent pays at least for its own renewal and that the more valuable patents will be renewed for a longer time, conditional on technology-specific differences in knowledge obsolescence.
Our conversations with university technology transfer managers provide support to this choice, as we were told that university administrations never renew patents that have not been licensed. They may sometimes go for selling them, after first filing, to companies interested in developing or commercializing the invention, which will then pay for the examination (if not yet completed) and the renewal fees. To the extent that this actually occurs, we judge that the university has effectively exploited the patent. It remains the possibility that firms that own or buy academic patents extract profits by keeping them sleeping. Notice however that distinguishing patents that are kept sleeping for strategic reasons from those that are renewed, but not exploited—in the wake of some possible exploitation—is rather difficult (Torrisi et al., 2016; Walsh et al., 2016).
One limitation of using Renewals as proxy of Patent Value is that the measure is applicable only to granted patents (rejected or withdrawn applications cannot be renewed), which forces us to reduce our patent sample when using it.15
As an alternative measure of Patent Value, we consider the size of the simple patent families to which our academic patents belong to (Patent Family), that is, the number of patent documents (mostly, applications) referring to the same invention across different national patent offices as well as within each office. More precisely, we use data for DOCDB patent families as defined by the EPO (Martínez, 2011). The size of the family increases with the number of countries the original (priority) patent is extended to and gets even larger if the patent is finally granted in more than one country (as this further increases the documentation entering the family). Here the logic is that only valuable patents justify paying for application, examination, and granting fees in different countries.
Coming to the Technological Importance of the invention, we measure it by means of two complementary citation-based indicators, both derived from the patent literature (for a recent survey, see Jaffe and de Rassenfosse, 2016).
First, we follow Henderson et al. (1998) in stating that an invention is the more important the more it stimulates and facilitates further inventions, as reflected by the number of its forward citations. We acknowledge, however, that forward citations may be affected by not only the characteristics of the invention but also by what we described as Exploitation and wish to keep distinct. In fact, an active assignee may generate a number of follow-up citing patents both by developing the invention internally and by licensing it actively. However, as suggested by Lanjouw and Schankerman (2004), this confusion is more likely to arise in the early years after the filing of the focal patent, rather than over the long run. With the passing of time, more and more patent applications are filed, which cite the focal patent but originate from research activities unrelated to the commercialization efforts of the focal patent’s assignee. Hence, we choose to count only the forward citations occurring at least 3 years after the academic patent’s priority date (Long-Term Citations). Notice that we consider both the citations received by the academic patents in our sample and by the other patents in their families.16
Coming at last to Exploitation, we face the difficulty of having no proxy to rely on, the assignees’ development and commercialization efforts, and their effectiveness, being unobservable. Hence, we content ourselves of a residual measurement, based on comparing the Patent Value of university-owned and firm-owned academic patents, after controlling for the Technological Importance and other characteristics of the patents. Based on our view of Patent Value being determined both by the Technological Importance and the effectiveness of Exploitation, we can then attribute any estimated difference to differences in the latter.
Figure 3 shows the average values of our measures of Technological Importance and Patent Value, for university-owned and firm-owned patents. We observe that the Technological Importance of the university-owned inventions is lower with respect to firm-owned inventions, both in terms of Long-Term Citations and Originality. Similarly, we observe that the Patent Value of the university-owned patents is lower, both in terms of Renewals and Family Size.
In our empirical analysis, we proceed in two steps. First, we estimate the probability of an academic patent to be assigned to a university rather than a firm, as a function of its Technological Importance and several controls. Second, based on such regressions, we match the university-owned academic patents to a control sample of firm-owned ones having the similar Technological Importance, and compare the mean Patent Value of the two samples.
3.2 Estimation strategy
The binary dependent variable in Equation (1), University, takes value 1 when the first applicant of the academic patent is a university and 0 when it is a firm. By first applicant we mean the organization that files the priority application at the patent office, regardless of whether it goes all the way to paying for international extensions or sell its rights to a further party right after filing. It is the first applicant who takes the initiative of seeking IP protection for the invention and eventually sells it as part of its exploitation strategy. Conversations with Italian TTO officers indicate that Italian universities often take the initiative of filing an application at the Italian office, immediately seeking for perspective IP buyers willing to pay for international extension.
The variable of interest in Equation (1) is the Technological Importance (TI) of the invention, as measured by both Long-Term Citations and Originality (see Section 3.1). In the absence of a clear causality nexus between TI and University, we cautiously interpret our results as correlations. Still, it is worth remarking that causality issues do not originate from a problem with our measurement of TI (the Originality of the invention is clearly antecedent to the application, and Long-Term Citations are reasonably unaffected by the type of ownership). Instead, they are related to the timing of the decisions concerning ownership that, as discussed in the previous section, may occur both before and after the moment when the importance of the invention is observed. In particular, although citations do occur after the patent assignment, along the patent life, they indicate the technological importance of the invention at the time of the filing. This is confirmed by Hall (2005), who found market value premia being highly correlated with future patent citations.
As for the controls, TC refers to the characteristics of the inventors’ team (the inventors named on the patent applications), whereas IP indicates any change affecting IP management at the university level, and X includes all further controls such as technology, year, and university fixed effects.
In particular, TC includes:
Team Size, which is the number of inventors listed on the patent application. Larger teams are associated to larger research projects, which in turn may require more institutional involvement on the part of the university. This increases the university’s control over the resulting IP, irrespective of any industry’s involvement in the project, and makes it more difficult for inventors to bypass the administration. Hence, we expect Team Size to be positively correlated with university ownership.
Share of Faculty in Team is the percentage of inventors in the team who appeared as tenured faculty members at the priority date of the patent. When equal to one, it signals that no industry researcher was part of the inventor team. This suggests no or little reliance on private sponsorship for the research leading to the invention, which increases the probability of university ownership.17
Full Professor is a dummy variable taking value equal to one if at least one of the inventors in the team was full professor at the time of the patent filing date. We suppose that full professors are more likely to retain IP control and pass it on eventually to firms, thanks to their stronger bargaining position with respect to the administration. We thus expect it to be negatively correlated to university ownership.
Age is the age of the oldest academic inventor among those listed on the focal patent application. We suppose that older researchers have more bargaining power than younger ones, face to the university administration. Moreover, they might have more contacts with firms potentially interested in their inventions. Finally, they may belong to a generation used to handling IP in the absence of TTOs or any explicit interest by university administration and retain a preference for acting autonomously.
Patenting Experience is the maximum number of patent applications filed by the inventors in the team, academic or not, and it reflects the patenting experience of the team of inventors. Similarly to Age, we expect experienced teams to have more chances to find a firm interested in their invention, and bypass the university, or to have a firm involved from the beginning in the inventive process or at least closely monitoring it.
As for IP, we consider two dummies:
IP Regulation takes value one if, at the priority date of the patent, the academic inventors’ university of affiliation had already adopted a formal IP regulation concerning its faculty’s inventions. In case the inventors belong to different universities, IP Regulation takes value one if at least one of such universities had introduced an IP regulation.
TTO takes value one if, at the priority date of the patent, the academic inventors’ university of affiliation had already set up a TTO. In case the inventors belong to different universities, for the variable to take value one, it suffices that at least one university had a TTO.18
For both IP Regulation and TTO, we expect a positive impact on the dependent variable, but also some correlation with university fixed effects (to the extent that universities with more aggressive IP strategies may also be the first ones to introduce regulations and tech-transfer offices).
We further control for a contextual variable that captures the environmental conditions in which universities operate, namely, the R&D intensity of local companies, from which academic inventors are most likely to obtain funds in exchange of IP rights. This is measured by the ratio between private and public R&D expenditures in the region of the inventors’ university of affiliation, at the priority date of the patent (R&D Business/R&D Public).19
Finally, we introduce fixed effects to control for the priority year, for the technological classes of the patent (30 classes based on the classification by Schmoch, 2008), and for the universities of affiliation of the inventors. Notice that year fixed effects are particularly relevant for controlling for the impact of the introduction of the professor’s privilege after 2001. To the same end, as an alternative specification, we replace the priority year fixed effects with the year trend (Priority Year) and a Professor’s Privilege dummy that takes value one for patents with priority year after 2001, so as to capture more distinctively the impact of the introduction of the professor’s privilege.
After estimating Equation (1), we move on to investigate Exploitation, by comparing the value of university-owned and firm-owned patents filed in the same years and in similar technological classes, with similar characteristics in terms of IP regulation and inventor teams, as well as, most importantly, with similar Technological Importance.
To do so, we proceed as follows:
First, we consider all the university-owned granted patents in our original sample (269 patents out of the original 700 patent applications). This restriction is necessary, as explained in the previous section, because only granted patents can be renewed.
Second, we build a matched sample of firm-owned granted patents with the same characteristics as the sample of university-owned patents. To do so, we apply propensity score matching (PSM), by proceeding in two steps.
In the first step, for each university-owned patent, we create a pool of firm-owned potential controls with the same priority year and technological macro-class (control pool). We impose the same priority year for a university-owned patent and its control so as to compare patents that have the same achievable maximum value of Renewals and Family Size, our measures for Patent Value.20 Similarly, we impose to the university-owned patent and its control to belong to the same technological macro-class owing to the class-specific propensity of reviewing and extending patent protection to multiple countries. For instance, renewals until the patent expiration are very common in pharmaceuticals and biotechnology, but much less so in electronics.21
In the second step, based on the estimated coefficients of Equation (1), we calculate the predicted probability of each focal and control patent to be university- or firm-owned, according to its characteristics; we then select from within each control pool the firm-owned patent with the closest propensity score to its university-owned equivalent.
After producing our control sample of firm-owned patents, we test if the average Patent Value in the two samples is statistically different. Having compared granted patents with similar Technological Importance, applied in the same year and in the same field, we attribute any difference to Exploitation.
4. Results
4.1 Main results
Tables 1 and 2 report, respectively, the descriptive statistics and estimation results for Equation (1). Specifications in Columns 1–3 of Table 2 are the same, except for the technology and university fixed effects, which are introduced in a stepwise fashion. In Column 4, we replace the priority year fixed effects with the year trend (Priority Year) and the Professor’s Privilege dummy. All columns report estimated logit coefficients.
. | N . | Min . | Max . | p50 . | Mean . | SD . |
---|---|---|---|---|---|---|
Dependent variable | ||||||
University | 2910 | 0 | 1 | 0 | 0.24 | 0.43 |
Independent variables | ||||||
TI | ||||||
Long-Term Citations | 2910 | 0 | 13 | 0 | 0.30 | 0.96 |
Originality | 2910 | 0 | 0.97 | 0.76 | 0.65 | 0.29 |
TC | ||||||
Share of Faculty in the Team | 2910 | 0.04 | 1 | 0.5 | 0.54 | 0.29 |
Team Size | 2910 | 1 | 49 | 3 | 3.89 | 2.53 |
Full Professorship | 2910 | 0 | 1 | 1 | 0.58 | 0.49 |
Age | 2910 | 28 | 82 | 54 | 53.89 | 9.96 |
Patenting Experience | 2910 | 1 | 40 | 2 | 4.17 | 4.78 |
IP | ||||||
IP Regulation | 2910 | 0 | 1 | 1 | 0.76 | 0.43 |
TTO | 2910 | 0 | 1 | 1 | 0.53 | 0.50 |
Year | 2910 | 1997 | 2009 | 2003 | 2003 | 3.43 |
R&D Business/R&D Public | 2910 | 0 | 4.73 | 1.12 | 1.59 | 1.18 |
. | N . | Min . | Max . | p50 . | Mean . | SD . |
---|---|---|---|---|---|---|
Dependent variable | ||||||
University | 2910 | 0 | 1 | 0 | 0.24 | 0.43 |
Independent variables | ||||||
TI | ||||||
Long-Term Citations | 2910 | 0 | 13 | 0 | 0.30 | 0.96 |
Originality | 2910 | 0 | 0.97 | 0.76 | 0.65 | 0.29 |
TC | ||||||
Share of Faculty in the Team | 2910 | 0.04 | 1 | 0.5 | 0.54 | 0.29 |
Team Size | 2910 | 1 | 49 | 3 | 3.89 | 2.53 |
Full Professorship | 2910 | 0 | 1 | 1 | 0.58 | 0.49 |
Age | 2910 | 28 | 82 | 54 | 53.89 | 9.96 |
Patenting Experience | 2910 | 1 | 40 | 2 | 4.17 | 4.78 |
IP | ||||||
IP Regulation | 2910 | 0 | 1 | 1 | 0.76 | 0.43 |
TTO | 2910 | 0 | 1 | 1 | 0.53 | 0.50 |
Year | 2910 | 1997 | 2009 | 2003 | 2003 | 3.43 |
R&D Business/R&D Public | 2910 | 0 | 4.73 | 1.12 | 1.59 | 1.18 |
. | N . | Min . | Max . | p50 . | Mean . | SD . |
---|---|---|---|---|---|---|
Dependent variable | ||||||
University | 2910 | 0 | 1 | 0 | 0.24 | 0.43 |
Independent variables | ||||||
TI | ||||||
Long-Term Citations | 2910 | 0 | 13 | 0 | 0.30 | 0.96 |
Originality | 2910 | 0 | 0.97 | 0.76 | 0.65 | 0.29 |
TC | ||||||
Share of Faculty in the Team | 2910 | 0.04 | 1 | 0.5 | 0.54 | 0.29 |
Team Size | 2910 | 1 | 49 | 3 | 3.89 | 2.53 |
Full Professorship | 2910 | 0 | 1 | 1 | 0.58 | 0.49 |
Age | 2910 | 28 | 82 | 54 | 53.89 | 9.96 |
Patenting Experience | 2910 | 1 | 40 | 2 | 4.17 | 4.78 |
IP | ||||||
IP Regulation | 2910 | 0 | 1 | 1 | 0.76 | 0.43 |
TTO | 2910 | 0 | 1 | 1 | 0.53 | 0.50 |
Year | 2910 | 1997 | 2009 | 2003 | 2003 | 3.43 |
R&D Business/R&D Public | 2910 | 0 | 4.73 | 1.12 | 1.59 | 1.18 |
. | N . | Min . | Max . | p50 . | Mean . | SD . |
---|---|---|---|---|---|---|
Dependent variable | ||||||
University | 2910 | 0 | 1 | 0 | 0.24 | 0.43 |
Independent variables | ||||||
TI | ||||||
Long-Term Citations | 2910 | 0 | 13 | 0 | 0.30 | 0.96 |
Originality | 2910 | 0 | 0.97 | 0.76 | 0.65 | 0.29 |
TC | ||||||
Share of Faculty in the Team | 2910 | 0.04 | 1 | 0.5 | 0.54 | 0.29 |
Team Size | 2910 | 1 | 49 | 3 | 3.89 | 2.53 |
Full Professorship | 2910 | 0 | 1 | 1 | 0.58 | 0.49 |
Age | 2910 | 28 | 82 | 54 | 53.89 | 9.96 |
Patenting Experience | 2910 | 1 | 40 | 2 | 4.17 | 4.78 |
IP | ||||||
IP Regulation | 2910 | 0 | 1 | 1 | 0.76 | 0.43 |
TTO | 2910 | 0 | 1 | 1 | 0.53 | 0.50 |
Year | 2910 | 1997 | 2009 | 2003 | 2003 | 3.43 |
R&D Business/R&D Public | 2910 | 0 | 4.73 | 1.12 | 1.59 | 1.18 |
. | (1) . | (2) . | (3) . | (4) . |
---|---|---|---|---|
Long-Term Citations | −0.245*** | −0.193*** | −0.233*** | −0.231*** |
(0.0779) | (0.0680) | (0.0813) | (0.0788) | |
Originality | −0.557*** | −0.539*** | −0.564*** | −0.569*** |
(0.159) | (0.174) | (0.189) | (0.188) | |
Share of Faculty in the Team | 2.679*** | 2.709*** | 2.762*** | 2.753*** |
(0.240) | (0.234) | (0.274) | (0.273) | |
Team Size | 0.145*** | 0.135*** | 0.109*** | 0.110*** |
(0.0455) | (0.0389) | (0.0410) | (0.0418) | |
Full Professorship | 0.577*** | 0.544*** | 0.542*** | 0.549*** |
(0.125) | (0.129) | (0.143) | (0.141) | |
Age | −0.0158*** | −0.0139** | −0.0130* | −0.0134* |
(0.00600) | (0.00638) | (0.00718) | (0.00713) | |
Patenting Experience | −0.127*** | −0.123*** | −0.125*** | −0.125*** |
(0.0192) | (0.0202) | (0.0217) | (0.0216) | |
IP Regulation | 0.382** | 0.596*** | 0.381 | 0.379 |
(0.158) | (0.171) | (0.238) | (0.231) | |
TTO | 0.374*** | 0.422*** | 0.454** | 0.480** |
(0.135) | (0.144) | (0.198) | (0.194) | |
R&D Business/R&D Public | −0.0552 | −0.0939** | −0.114 | −0.164 |
(0.0427) | (0.0454) | (0.142) | (0.137) | |
Professor's Privilege | 0.195 | |||
(0.218) | ||||
Priority Year | 0.175*** | |||
(0.0367) | ||||
Constant | −3.780*** | −3.963*** | −4.338*** | −354.6*** |
(0.488) | (0.538) | (0.606) | (73.43) | |
Observations | 2910 | 2910 | 2910 | 2910 |
Pseudo R2 | 0.19 | 0.25 | 0.32 | 0.31 |
Year | Yes | Yes | Yes | No |
Technological Field | No | Yes | Yes | Yes |
University | No | No | Yes | Yes |
. | (1) . | (2) . | (3) . | (4) . |
---|---|---|---|---|
Long-Term Citations | −0.245*** | −0.193*** | −0.233*** | −0.231*** |
(0.0779) | (0.0680) | (0.0813) | (0.0788) | |
Originality | −0.557*** | −0.539*** | −0.564*** | −0.569*** |
(0.159) | (0.174) | (0.189) | (0.188) | |
Share of Faculty in the Team | 2.679*** | 2.709*** | 2.762*** | 2.753*** |
(0.240) | (0.234) | (0.274) | (0.273) | |
Team Size | 0.145*** | 0.135*** | 0.109*** | 0.110*** |
(0.0455) | (0.0389) | (0.0410) | (0.0418) | |
Full Professorship | 0.577*** | 0.544*** | 0.542*** | 0.549*** |
(0.125) | (0.129) | (0.143) | (0.141) | |
Age | −0.0158*** | −0.0139** | −0.0130* | −0.0134* |
(0.00600) | (0.00638) | (0.00718) | (0.00713) | |
Patenting Experience | −0.127*** | −0.123*** | −0.125*** | −0.125*** |
(0.0192) | (0.0202) | (0.0217) | (0.0216) | |
IP Regulation | 0.382** | 0.596*** | 0.381 | 0.379 |
(0.158) | (0.171) | (0.238) | (0.231) | |
TTO | 0.374*** | 0.422*** | 0.454** | 0.480** |
(0.135) | (0.144) | (0.198) | (0.194) | |
R&D Business/R&D Public | −0.0552 | −0.0939** | −0.114 | −0.164 |
(0.0427) | (0.0454) | (0.142) | (0.137) | |
Professor's Privilege | 0.195 | |||
(0.218) | ||||
Priority Year | 0.175*** | |||
(0.0367) | ||||
Constant | −3.780*** | −3.963*** | −4.338*** | −354.6*** |
(0.488) | (0.538) | (0.606) | (73.43) | |
Observations | 2910 | 2910 | 2910 | 2910 |
Pseudo R2 | 0.19 | 0.25 | 0.32 | 0.31 |
Year | Yes | Yes | Yes | No |
Technological Field | No | Yes | Yes | Yes |
University | No | No | Yes | Yes |
The table reports estimated coefficients (robust standard errors in parentheses). *90% significance; **95% significance; ***99% significance.
The sample includes only academic patents owned by either universities or firms.
. | (1) . | (2) . | (3) . | (4) . |
---|---|---|---|---|
Long-Term Citations | −0.245*** | −0.193*** | −0.233*** | −0.231*** |
(0.0779) | (0.0680) | (0.0813) | (0.0788) | |
Originality | −0.557*** | −0.539*** | −0.564*** | −0.569*** |
(0.159) | (0.174) | (0.189) | (0.188) | |
Share of Faculty in the Team | 2.679*** | 2.709*** | 2.762*** | 2.753*** |
(0.240) | (0.234) | (0.274) | (0.273) | |
Team Size | 0.145*** | 0.135*** | 0.109*** | 0.110*** |
(0.0455) | (0.0389) | (0.0410) | (0.0418) | |
Full Professorship | 0.577*** | 0.544*** | 0.542*** | 0.549*** |
(0.125) | (0.129) | (0.143) | (0.141) | |
Age | −0.0158*** | −0.0139** | −0.0130* | −0.0134* |
(0.00600) | (0.00638) | (0.00718) | (0.00713) | |
Patenting Experience | −0.127*** | −0.123*** | −0.125*** | −0.125*** |
(0.0192) | (0.0202) | (0.0217) | (0.0216) | |
IP Regulation | 0.382** | 0.596*** | 0.381 | 0.379 |
(0.158) | (0.171) | (0.238) | (0.231) | |
TTO | 0.374*** | 0.422*** | 0.454** | 0.480** |
(0.135) | (0.144) | (0.198) | (0.194) | |
R&D Business/R&D Public | −0.0552 | −0.0939** | −0.114 | −0.164 |
(0.0427) | (0.0454) | (0.142) | (0.137) | |
Professor's Privilege | 0.195 | |||
(0.218) | ||||
Priority Year | 0.175*** | |||
(0.0367) | ||||
Constant | −3.780*** | −3.963*** | −4.338*** | −354.6*** |
(0.488) | (0.538) | (0.606) | (73.43) | |
Observations | 2910 | 2910 | 2910 | 2910 |
Pseudo R2 | 0.19 | 0.25 | 0.32 | 0.31 |
Year | Yes | Yes | Yes | No |
Technological Field | No | Yes | Yes | Yes |
University | No | No | Yes | Yes |
. | (1) . | (2) . | (3) . | (4) . |
---|---|---|---|---|
Long-Term Citations | −0.245*** | −0.193*** | −0.233*** | −0.231*** |
(0.0779) | (0.0680) | (0.0813) | (0.0788) | |
Originality | −0.557*** | −0.539*** | −0.564*** | −0.569*** |
(0.159) | (0.174) | (0.189) | (0.188) | |
Share of Faculty in the Team | 2.679*** | 2.709*** | 2.762*** | 2.753*** |
(0.240) | (0.234) | (0.274) | (0.273) | |
Team Size | 0.145*** | 0.135*** | 0.109*** | 0.110*** |
(0.0455) | (0.0389) | (0.0410) | (0.0418) | |
Full Professorship | 0.577*** | 0.544*** | 0.542*** | 0.549*** |
(0.125) | (0.129) | (0.143) | (0.141) | |
Age | −0.0158*** | −0.0139** | −0.0130* | −0.0134* |
(0.00600) | (0.00638) | (0.00718) | (0.00713) | |
Patenting Experience | −0.127*** | −0.123*** | −0.125*** | −0.125*** |
(0.0192) | (0.0202) | (0.0217) | (0.0216) | |
IP Regulation | 0.382** | 0.596*** | 0.381 | 0.379 |
(0.158) | (0.171) | (0.238) | (0.231) | |
TTO | 0.374*** | 0.422*** | 0.454** | 0.480** |
(0.135) | (0.144) | (0.198) | (0.194) | |
R&D Business/R&D Public | −0.0552 | −0.0939** | −0.114 | −0.164 |
(0.0427) | (0.0454) | (0.142) | (0.137) | |
Professor's Privilege | 0.195 | |||
(0.218) | ||||
Priority Year | 0.175*** | |||
(0.0367) | ||||
Constant | −3.780*** | −3.963*** | −4.338*** | −354.6*** |
(0.488) | (0.538) | (0.606) | (73.43) | |
Observations | 2910 | 2910 | 2910 | 2910 |
Pseudo R2 | 0.19 | 0.25 | 0.32 | 0.31 |
Year | Yes | Yes | Yes | No |
Technological Field | No | Yes | Yes | Yes |
University | No | No | Yes | Yes |
The table reports estimated coefficients (robust standard errors in parentheses). *90% significance; **95% significance; ***99% significance.
The sample includes only academic patents owned by either universities or firms.
As expected, we find that the Technological Importance of the inventions is negatively correlated with university ownership. Both the coefficients for Long-Term Citations and Originality are negative and significant, with rather stable values across all columns. Our preferred specification is that of Column 3: there, the value of the coefficient for Long-Term Citations is −0.23 and statistically significant. This implies an odds ratio equal to 0.79 (e−0.23), namely, a 21% decrease in the probability of university ownership, relative to firm ownership, for any extra citation.
The negative sign for the estimated coefficients of Age and Patenting Experience suggests that we successfully control for any lack of disclosure on the part of older and more experienced academic inventors. However, this is not the case for Full Professor, whose coefficient we hypothesized to be negative, but turns out positive. One possible explanation is that, after controlling for age and experience, senior faculty members use their bargaining power not to retain IP but to offload it, when of poor quality, to their administrations, which will bear the cost of patenting it. These patents may increase the professors’ prestige or visibility and be a useful signal for the university as well.
As discussed in Section 2, the ownership of academic patents may be largely determined by the contractual conditions underlying the research behind the invention, with business sponsorship both decreasing the odds of university ownership and originating more technologically important inventions. The positive and significant coefficient of Share of Faculty in Team suggests that universities indeed are the more likely to own the patents the less industry is involved in the research project (the higher the percentage of academic inventors in the team).
Our results also show that patents involving larger teams are associated to a higher probability of university ownership. In particular, the coefficient for Team Size in Column 3 indicates an odds ratio of 1.11, that is, an 11% increase in the probability of university ownership, relative to firm ownership, for any additional inventor in the team.
The estimated coefficients for IP regime changes at the university level have the expected signs and are rather stable across specifications. IP Regulation, however, is more sensitive to the introduction of university fixed effects and becomes not significant in Columns 3 and 4. As for the TTO dummy, results in Column 3 imply that its presence increases by about 58% the probability of university ownership, relative to the probability of firm ownership.
The specification of Column 4 controls more directly for the effect of the introduction of the professor’s privilege in 2001, as distinct from any visible ownership trend. We obtain a positive and significant effect of the time trend variable, with an odds ratio equal to 1.19 for every additional year. This confirms the existence of a positive trend toward university ownership. In contrast, the estimated coefficient for Professor’s Privilege is never significant, which confirms the irrelevance of this legislative change. Figure 4 plots the predicted probability of university ownership by priority year of the patent: a positive trend is clearly visible, which is not altered by the introduction of the professor’s privilege in 2001 and its subsequent reformulation in 2005.
For the estimation of the Exploitation effect through the PSM method described in Section 3.2, we rely on the estimated coefficients in Column 3 of Table 2.
Table 3 reports the average values of the characteristics of the 269 university-owned patents, the 1011 firm-owned patents from which we draw our control sample, and the 269 firm-owned patents that we select as the control group with the PSM method. We notice that the first and second groups are significantly different along different dimensions with the exception of the variable R&D Business/R&D Public (Columns 3 and 4). Instead, when we compare the first and third groups, all differences become nonsignificant, with the only exception of the Team Size variable (Columns 6 and 7). In particular, the two matched samples show similar values for Long-Term Citations and Originality, which goes against the possibility of a selection bias due to differences in Technological Importance.
. | University-owned patents (269 obs.) . | Firm-owned patents before the match (1011 obs.) . | Firm-owned patents after the match (269 obs.) . | ||||
---|---|---|---|---|---|---|---|
(1) . | (2) . | (3) . | (4) . | (5) . | (6) . | (7) . | |
Average . | Average . | Difference . | P-value . | Average . | Difference . | P-value . | |
TI | (1)-(2) | (1)-(5) | |||||
Long-Term Citations | 0.14 | 0.54 | −0.39*** | 0.00 | 0.16 | −0.02 | 0.75 |
Originality | 0.58 | 0.64 | −0.05** | 0.01 | 0.61 | −0.03 | 0.34 |
TC | |||||||
Share of Faculty in the Team | 0.71 | 0.53 | 0.18*** | 0.00 | 0.69 | 0.02 | 0.71 |
Team Size | 3.91 | 3.66 | 0.25* | 0.06 | 3.58 | 0.34* | 0.07 |
Full Professorship | 0.67 | 0.5 | 0.17*** | 0.00 | 0.68 | −0.01 | 0.58 |
Age | 55.57 | 53.09 | 2.48*** | 0.00 | 55.69 | −0.12 | 0.89 |
Patenting Experience | 3.14 | 4.08 | −0.95*** | 0.00 | 3.16 | −0.01 | 0.97 |
IP | |||||||
IP Regulation | 0.86 | 0.68 | 0.18*** | 0.00 | 0.85 | 0.00 | 0.90 |
TTO | 0.68 | 0.4 | 0.28*** | 0.00 | 0.66 | 0.01 | 0.71 |
R&D Business / R&D Public | 1.63 | 1.58 | 0.05 | 0.56 | 1.70 | −0.07 | 0.51 |
. | University-owned patents (269 obs.) . | Firm-owned patents before the match (1011 obs.) . | Firm-owned patents after the match (269 obs.) . | ||||
---|---|---|---|---|---|---|---|
(1) . | (2) . | (3) . | (4) . | (5) . | (6) . | (7) . | |
Average . | Average . | Difference . | P-value . | Average . | Difference . | P-value . | |
TI | (1)-(2) | (1)-(5) | |||||
Long-Term Citations | 0.14 | 0.54 | −0.39*** | 0.00 | 0.16 | −0.02 | 0.75 |
Originality | 0.58 | 0.64 | −0.05** | 0.01 | 0.61 | −0.03 | 0.34 |
TC | |||||||
Share of Faculty in the Team | 0.71 | 0.53 | 0.18*** | 0.00 | 0.69 | 0.02 | 0.71 |
Team Size | 3.91 | 3.66 | 0.25* | 0.06 | 3.58 | 0.34* | 0.07 |
Full Professorship | 0.67 | 0.5 | 0.17*** | 0.00 | 0.68 | −0.01 | 0.58 |
Age | 55.57 | 53.09 | 2.48*** | 0.00 | 55.69 | −0.12 | 0.89 |
Patenting Experience | 3.14 | 4.08 | −0.95*** | 0.00 | 3.16 | −0.01 | 0.97 |
IP | |||||||
IP Regulation | 0.86 | 0.68 | 0.18*** | 0.00 | 0.85 | 0.00 | 0.90 |
TTO | 0.68 | 0.4 | 0.28*** | 0.00 | 0.66 | 0.01 | 0.71 |
R&D Business / R&D Public | 1.63 | 1.58 | 0.05 | 0.56 | 1.70 | −0.07 | 0.51 |
*90% significance; **95% significance; ***99% significance.
. | University-owned patents (269 obs.) . | Firm-owned patents before the match (1011 obs.) . | Firm-owned patents after the match (269 obs.) . | ||||
---|---|---|---|---|---|---|---|
(1) . | (2) . | (3) . | (4) . | (5) . | (6) . | (7) . | |
Average . | Average . | Difference . | P-value . | Average . | Difference . | P-value . | |
TI | (1)-(2) | (1)-(5) | |||||
Long-Term Citations | 0.14 | 0.54 | −0.39*** | 0.00 | 0.16 | −0.02 | 0.75 |
Originality | 0.58 | 0.64 | −0.05** | 0.01 | 0.61 | −0.03 | 0.34 |
TC | |||||||
Share of Faculty in the Team | 0.71 | 0.53 | 0.18*** | 0.00 | 0.69 | 0.02 | 0.71 |
Team Size | 3.91 | 3.66 | 0.25* | 0.06 | 3.58 | 0.34* | 0.07 |
Full Professorship | 0.67 | 0.5 | 0.17*** | 0.00 | 0.68 | −0.01 | 0.58 |
Age | 55.57 | 53.09 | 2.48*** | 0.00 | 55.69 | −0.12 | 0.89 |
Patenting Experience | 3.14 | 4.08 | −0.95*** | 0.00 | 3.16 | −0.01 | 0.97 |
IP | |||||||
IP Regulation | 0.86 | 0.68 | 0.18*** | 0.00 | 0.85 | 0.00 | 0.90 |
TTO | 0.68 | 0.4 | 0.28*** | 0.00 | 0.66 | 0.01 | 0.71 |
R&D Business / R&D Public | 1.63 | 1.58 | 0.05 | 0.56 | 1.70 | −0.07 | 0.51 |
. | University-owned patents (269 obs.) . | Firm-owned patents before the match (1011 obs.) . | Firm-owned patents after the match (269 obs.) . | ||||
---|---|---|---|---|---|---|---|
(1) . | (2) . | (3) . | (4) . | (5) . | (6) . | (7) . | |
Average . | Average . | Difference . | P-value . | Average . | Difference . | P-value . | |
TI | (1)-(2) | (1)-(5) | |||||
Long-Term Citations | 0.14 | 0.54 | −0.39*** | 0.00 | 0.16 | −0.02 | 0.75 |
Originality | 0.58 | 0.64 | −0.05** | 0.01 | 0.61 | −0.03 | 0.34 |
TC | |||||||
Share of Faculty in the Team | 0.71 | 0.53 | 0.18*** | 0.00 | 0.69 | 0.02 | 0.71 |
Team Size | 3.91 | 3.66 | 0.25* | 0.06 | 3.58 | 0.34* | 0.07 |
Full Professorship | 0.67 | 0.5 | 0.17*** | 0.00 | 0.68 | −0.01 | 0.58 |
Age | 55.57 | 53.09 | 2.48*** | 0.00 | 55.69 | −0.12 | 0.89 |
Patenting Experience | 3.14 | 4.08 | −0.95*** | 0.00 | 3.16 | −0.01 | 0.97 |
IP | |||||||
IP Regulation | 0.86 | 0.68 | 0.18*** | 0.00 | 0.85 | 0.00 | 0.90 |
TTO | 0.68 | 0.4 | 0.28*** | 0.00 | 0.66 | 0.01 | 0.71 |
R&D Business / R&D Public | 1.63 | 1.58 | 0.05 | 0.56 | 1.70 | −0.07 | 0.51 |
*90% significance; **95% significance; ***99% significance.
The methodology used to build the control sample of firm-owned academic patents is described in Section 3.2.
Table 4 shows the average Patent Value for the university-owned and the matched firm-owned patents, as measured by Renewals and Family Size. We find that, conditional on measured Technological Importance (and other observables characteristics of university ownership), universities exploit patents less successful: university-owned patents are associated to 0.58 less Renewals and 2.25 less countries in the family (Family Size).
. | University-owned patents (269 obs.) . | Firm-owned patents (269 obs.) . | Difference . | P-value . |
---|---|---|---|---|
Renewals | 7.14 | 7.72 | −0.58** | 0.01 |
Family Size | 6.96 | 9.21 | −2.25*** | 0.00 |
. | University-owned patents (269 obs.) . | Firm-owned patents (269 obs.) . | Difference . | P-value . |
---|---|---|---|---|
Renewals | 7.14 | 7.72 | −0.58** | 0.01 |
Family Size | 6.96 | 9.21 | −2.25*** | 0.00 |
*90% significance; **95% significance; ***99% significance.
. | University-owned patents (269 obs.) . | Firm-owned patents (269 obs.) . | Difference . | P-value . |
---|---|---|---|---|
Renewals | 7.14 | 7.72 | −0.58** | 0.01 |
Family Size | 6.96 | 9.21 | −2.25*** | 0.00 |
. | University-owned patents (269 obs.) . | Firm-owned patents (269 obs.) . | Difference . | P-value . |
---|---|---|---|---|
Renewals | 7.14 | 7.72 | −0.58** | 0.01 |
Family Size | 6.96 | 9.21 | −2.25*** | 0.00 |
*90% significance; **95% significance; ***99% significance.
4.2 Further results and interpretation
According to the results presented so far, universities retain ownership of less important academic patents and exploit them less effectively than firms. These results require further investigation, for interpretation purposes. Concerning Technological Importance, and to the extent that at least in some cases the decision on ownership is taken after the production of inventions, there are two, nonmutually exclusive possibilities: first, the university administration may not evaluate correctly the importance of its faculty’s invention (or lack thereof); second, the administration decides on patenting irrespective of the invention’s importance, owing to its peculiar set of incentives. We may speculate that although the set of incentives has not changed much over the decade we consider, the university administrations’ grasp of Technological Importance should have improved on the basis of the experience acquired over the years (remember that Italian universities gained full autonomy only in 1996, which suggests that, in the early years covered by our study, direct IP management was still a new activity for many of them).
In Column 1 of Table 5, we exploit both longitudinal and cross-sectional variation in our data and calculate, for each university, the cumulative number of patents filed as first applicants (University Patenting Experience) at the time of the focal patent. We add this new variable to Equation (1) and we interact it with both TI indicators (Long-Term Citations and Originality). We find that University Patenting Experience attenuates the negative link between the probability of university ownership and the Technological Importance of the inventions, as witnessed by the positive and significant coefficient of the interaction of University Patenting Experience and Long-Term Citations. In the same table, the coefficient of the interaction of University Patenting Experience and Originality is also positive, but not significant.
. | (1) . | (2) . |
---|---|---|
Long-Term Citations | −0.696*** | −0.259** |
(0.252) | (0.105) | |
Originality | −0.641 | −0.821*** |
(0.473) | (0.283) | |
Share of Faculty in the Team | 2.831*** | 2.762*** |
(0.284) | (0.276) | |
Team Size | 0.111*** | 0.109*** |
(0.0425) | (0.0418) | |
Full Professorship | 0.522*** | 0.542*** |
(0.146) | (0.142) | |
Faculty's Age | −0.0117 | −0.0129* |
(0.00740) | (0.00717) | |
Faculty's Patenting Experience | −0.127*** | −0.129*** |
(0.0224) | (0.0225) | |
IP Regulation | 0.0589 | 0.370 |
(0.250) | (0.239) | |
TTO | 0.448** | 0.197 |
(0.202) | (0.297) | |
R&D Business/R&D Public | −0.200 | −0.132 |
(0.146) | (0.143) | |
log(University Patenting Experience) | 1.148*** | |
(0.183) | ||
Long-Term Citations × log(University Patenting Experience) | 0.236** | |
(0.103) | ||
Originality × log(University Patenting Experience) | 0.0229 | |
(0.169) | ||
Long-Term Citations × TTO | 0.0970 | |
(0.150) | ||
Originality × TTO | 0.409 | |
(0.366) | ||
Constant | −3.946*** | −4.148*** |
(0.646) | (0.621) | |
Observations | 2, 910 | 2, 910 |
Pseudo R2 | 0.34 | 0.32 |
Year | YES | YES |
Technological Field | YES | YES |
University | YES | YES |
. | (1) . | (2) . |
---|---|---|
Long-Term Citations | −0.696*** | −0.259** |
(0.252) | (0.105) | |
Originality | −0.641 | −0.821*** |
(0.473) | (0.283) | |
Share of Faculty in the Team | 2.831*** | 2.762*** |
(0.284) | (0.276) | |
Team Size | 0.111*** | 0.109*** |
(0.0425) | (0.0418) | |
Full Professorship | 0.522*** | 0.542*** |
(0.146) | (0.142) | |
Faculty's Age | −0.0117 | −0.0129* |
(0.00740) | (0.00717) | |
Faculty's Patenting Experience | −0.127*** | −0.129*** |
(0.0224) | (0.0225) | |
IP Regulation | 0.0589 | 0.370 |
(0.250) | (0.239) | |
TTO | 0.448** | 0.197 |
(0.202) | (0.297) | |
R&D Business/R&D Public | −0.200 | −0.132 |
(0.146) | (0.143) | |
log(University Patenting Experience) | 1.148*** | |
(0.183) | ||
Long-Term Citations × log(University Patenting Experience) | 0.236** | |
(0.103) | ||
Originality × log(University Patenting Experience) | 0.0229 | |
(0.169) | ||
Long-Term Citations × TTO | 0.0970 | |
(0.150) | ||
Originality × TTO | 0.409 | |
(0.366) | ||
Constant | −3.946*** | −4.148*** |
(0.646) | (0.621) | |
Observations | 2, 910 | 2, 910 |
Pseudo R2 | 0.34 | 0.32 |
Year | YES | YES |
Technological Field | YES | YES |
University | YES | YES |
The table reports estimated coefficients (robust standard errors in parentheses). *90% significance; **95% significance; ***99% significance.
The sample includes only academic patents owned by either universities or firms.
. | (1) . | (2) . |
---|---|---|
Long-Term Citations | −0.696*** | −0.259** |
(0.252) | (0.105) | |
Originality | −0.641 | −0.821*** |
(0.473) | (0.283) | |
Share of Faculty in the Team | 2.831*** | 2.762*** |
(0.284) | (0.276) | |
Team Size | 0.111*** | 0.109*** |
(0.0425) | (0.0418) | |
Full Professorship | 0.522*** | 0.542*** |
(0.146) | (0.142) | |
Faculty's Age | −0.0117 | −0.0129* |
(0.00740) | (0.00717) | |
Faculty's Patenting Experience | −0.127*** | −0.129*** |
(0.0224) | (0.0225) | |
IP Regulation | 0.0589 | 0.370 |
(0.250) | (0.239) | |
TTO | 0.448** | 0.197 |
(0.202) | (0.297) | |
R&D Business/R&D Public | −0.200 | −0.132 |
(0.146) | (0.143) | |
log(University Patenting Experience) | 1.148*** | |
(0.183) | ||
Long-Term Citations × log(University Patenting Experience) | 0.236** | |
(0.103) | ||
Originality × log(University Patenting Experience) | 0.0229 | |
(0.169) | ||
Long-Term Citations × TTO | 0.0970 | |
(0.150) | ||
Originality × TTO | 0.409 | |
(0.366) | ||
Constant | −3.946*** | −4.148*** |
(0.646) | (0.621) | |
Observations | 2, 910 | 2, 910 |
Pseudo R2 | 0.34 | 0.32 |
Year | YES | YES |
Technological Field | YES | YES |
University | YES | YES |
. | (1) . | (2) . |
---|---|---|
Long-Term Citations | −0.696*** | −0.259** |
(0.252) | (0.105) | |
Originality | −0.641 | −0.821*** |
(0.473) | (0.283) | |
Share of Faculty in the Team | 2.831*** | 2.762*** |
(0.284) | (0.276) | |
Team Size | 0.111*** | 0.109*** |
(0.0425) | (0.0418) | |
Full Professorship | 0.522*** | 0.542*** |
(0.146) | (0.142) | |
Faculty's Age | −0.0117 | −0.0129* |
(0.00740) | (0.00717) | |
Faculty's Patenting Experience | −0.127*** | −0.129*** |
(0.0224) | (0.0225) | |
IP Regulation | 0.0589 | 0.370 |
(0.250) | (0.239) | |
TTO | 0.448** | 0.197 |
(0.202) | (0.297) | |
R&D Business/R&D Public | −0.200 | −0.132 |
(0.146) | (0.143) | |
log(University Patenting Experience) | 1.148*** | |
(0.183) | ||
Long-Term Citations × log(University Patenting Experience) | 0.236** | |
(0.103) | ||
Originality × log(University Patenting Experience) | 0.0229 | |
(0.169) | ||
Long-Term Citations × TTO | 0.0970 | |
(0.150) | ||
Originality × TTO | 0.409 | |
(0.366) | ||
Constant | −3.946*** | −4.148*** |
(0.646) | (0.621) | |
Observations | 2, 910 | 2, 910 |
Pseudo R2 | 0.34 | 0.32 |
Year | YES | YES |
Technological Field | YES | YES |
University | YES | YES |
The table reports estimated coefficients (robust standard errors in parentheses). *90% significance; **95% significance; ***99% significance.
The sample includes only academic patents owned by either universities or firms.
We also consider the possibility that opening a TTO may compensate for the university administration’s lack of experience and increase the Technological Importance of the inventions retained for patenting, although we know that for Italian universities it may be difficult to staff it with an experienced IP manager. To this end, in Column 2 of Table 5, we interact the TTO dummy with Long-Term Citations and Originality in Equation (1). Neither interactions is statistically significant, meaning that the presence of a TTO per se did not improve the universities’ IP management of academic inventions, possibly because none or just a few were staffed with experienced personnel.
Experience may also matter with respect to Exploitation, to the extent that universities may learn how to better commercialize their academic patents by accumulating them. We investigate this possibility by classifying university-owned patents according to the university’s patenting experience at the time of their filing. In particular, we distinguish between high- and low-experience universities. The former are those having filed more than 14 patents (University Patenting Experience > 14) at the date of filing of the focal patent, where 14 is the median value of the cumulated number of patents filed by the universities in the group of university-owned patents. Notice that the same university may appear both in the high- and the low-experience sample, depending on the filing year of the focal patent.22
Based on this classification, we split our matched sample and calculate the average Patent Value for both university-owned and firm-owned patents, respectively, when universities are highly experienced (139 focal patents and 139 controls) and when they are not (130 focal patents and 130 controls). Results are reported in Table 6. We find that the difference between the average Patent Value of university- and firm-owned patents remains negative also when we restrict the sample to high-experience universities. Specifically, in such a case, university-owned patents receive 0.5 less renewals and have 1.45 less patent documents in their family. As for low-experience universities, university-owned patents have 0.67 less renewals and 3.09 less documents in the family.23
. | University-owned patents . | Firm-owned patents . | Difference . | P-value . |
---|---|---|---|---|
(1) . | (2) . | (1) − (2) . | ||
High Patenting Experience | (139 obs.) | (139 obs.) | ||
Average Renewals | 6.13 | 6.64 | −0.50 | 0.04 |
Average Family Size | 6.50 | 7.94 | −1.45 | 0.006 |
Low Patenting Experience | (130 obs.) | (130 obs.) | ||
Average Renewals | 8.22 | 8.89 | −0.67 | 0.05 |
Average Family Size | 7.46 | 10.55 | −3.09 | 0.00 |
University with TTO | (183 obs.) | (183 obs.) | ||
Average Renewals | 6.33 | 6.84 | −0.51 | 0.02 |
Average Family Size | 6.51 | 8.90 | −2.38 | 0.00 |
University without TTO | (86 obs.) | (86 obs.) | ||
Average Renewals | 8.87 | 9.61 | −0.74 | 0.07 |
Average Family Size | 7.93 | 9.87 | −1.94 | 0.03 |
. | University-owned patents . | Firm-owned patents . | Difference . | P-value . |
---|---|---|---|---|
(1) . | (2) . | (1) − (2) . | ||
High Patenting Experience | (139 obs.) | (139 obs.) | ||
Average Renewals | 6.13 | 6.64 | −0.50 | 0.04 |
Average Family Size | 6.50 | 7.94 | −1.45 | 0.006 |
Low Patenting Experience | (130 obs.) | (130 obs.) | ||
Average Renewals | 8.22 | 8.89 | −0.67 | 0.05 |
Average Family Size | 7.46 | 10.55 | −3.09 | 0.00 |
University with TTO | (183 obs.) | (183 obs.) | ||
Average Renewals | 6.33 | 6.84 | −0.51 | 0.02 |
Average Family Size | 6.51 | 8.90 | −2.38 | 0.00 |
University without TTO | (86 obs.) | (86 obs.) | ||
Average Renewals | 8.87 | 9.61 | −0.74 | 0.07 |
Average Family Size | 7.93 | 9.87 | −1.94 | 0.03 |
. | University-owned patents . | Firm-owned patents . | Difference . | P-value . |
---|---|---|---|---|
(1) . | (2) . | (1) − (2) . | ||
High Patenting Experience | (139 obs.) | (139 obs.) | ||
Average Renewals | 6.13 | 6.64 | −0.50 | 0.04 |
Average Family Size | 6.50 | 7.94 | −1.45 | 0.006 |
Low Patenting Experience | (130 obs.) | (130 obs.) | ||
Average Renewals | 8.22 | 8.89 | −0.67 | 0.05 |
Average Family Size | 7.46 | 10.55 | −3.09 | 0.00 |
University with TTO | (183 obs.) | (183 obs.) | ||
Average Renewals | 6.33 | 6.84 | −0.51 | 0.02 |
Average Family Size | 6.51 | 8.90 | −2.38 | 0.00 |
University without TTO | (86 obs.) | (86 obs.) | ||
Average Renewals | 8.87 | 9.61 | −0.74 | 0.07 |
Average Family Size | 7.93 | 9.87 | −1.94 | 0.03 |
. | University-owned patents . | Firm-owned patents . | Difference . | P-value . |
---|---|---|---|---|
(1) . | (2) . | (1) − (2) . | ||
High Patenting Experience | (139 obs.) | (139 obs.) | ||
Average Renewals | 6.13 | 6.64 | −0.50 | 0.04 |
Average Family Size | 6.50 | 7.94 | −1.45 | 0.006 |
Low Patenting Experience | (130 obs.) | (130 obs.) | ||
Average Renewals | 8.22 | 8.89 | −0.67 | 0.05 |
Average Family Size | 7.46 | 10.55 | −3.09 | 0.00 |
University with TTO | (183 obs.) | (183 obs.) | ||
Average Renewals | 6.33 | 6.84 | −0.51 | 0.02 |
Average Family Size | 6.51 | 8.90 | −2.38 | 0.00 |
University without TTO | (86 obs.) | (86 obs.) | ||
Average Renewals | 8.87 | 9.61 | −0.74 | 0.07 |
Average Family Size | 7.93 | 9.87 | −1.94 | 0.03 |
When we consider the absolute values of Renewals and Family Size, we notice that low-experience university-owned academic patents have, on average, more renewals and larger families than high-experience ones, and that a similar pattern is observed for the group of firm-owned control patents. This is owing to a composition effect: patents by low-experience universities are mainly concentrated in the first part of our observation period, whereas those by high-experience universities are more recent and, at the same time, have had less time to be renewed. For this reason, the averages in the samples of high- and low-experience universities cannot be compared.24
We perform a similar exercise for what concerns the presence of a TTO in the university at the time of patent filing, that is, we distinguish two groups of university-owned patents, namely, those filed when the university already had opened a TTO at the time of the patent filing and those filed in the absence of a TTO. For each group, we calculate the Patent Value difference with the firm-owned patents selected as controls. We find that the difference in the Patent Value persists even when we restrict the sample to university applicants with TTO university-owned, which suggests that TTOs do not improve Exploitation.
5. Discussion and conclusions
Over the past quarter century, university-owned patents have increased dramatically, both in absolute numbers and as a share of total academic patenting (patenting of academic inventions). The trend has been most striking in Europe, wherein many countries originally stood out for the high share of academic patents owned by firms and, to a lesser extent, individual inventors. Several studies have investigated the value of academic patents in relation to their ownership and often found university-owned ones to score modestly according to more than one metrics. In this article, based on a longitudinal sample of Italian academic patents, we have tried to deepen this evidence by investigating the reasons why the value of university-owned may be lower than that of firm-owned ones.
We have proposed a conceptual distinction between two sources of Patent Value, namely, the Technological Importance of the academic inventions appropriated through patent filing and the effectiveness of their Exploitation. We have found that, over the period 1997–2009, Italian universities have both filed patent applications for less important inventions and proved to be less effective than firms at exploiting the granted patents. Although the relationship between patent ownership and the importance of the inventions may run in both ways, our evidence is at least suggestive of some selection problems, by which universities somehow decide to file patents for less important academic inventions than those left to firms. Together with results on Exploitation, this suggests that Italian universities have several problems with IP management.
When looking deeper into these problems, we found that the university’s patenting experience with patent handling impacted in a limited way on the Technological Importance of filed patents and not at all on Exploitation. Nor do universities seem to do better in any dimension in presence of a TTO.
This leaves us with a tentative explanation based on differences between the system of incentives dictating universities’ and firms’ incentives to file patent applications over academic inventions. In particular, universities may value such patent applications for the signal they send to both private and public sponsors on their willingness to engage in the commercialization of their faculty’s research results. This may be a response to the increasing share of public funding distributed on a competitive basis, with the performance of universities being also assessed with technology transfer indicators, and the general pressure on universities to increase their patent portfolios. This residual explanation is certainly worth investigating further, one tool for doing so being possibly a mix of in-depth interviews to university administrators and extensive questionnaire survey of inventors.
From the policy viewpoint, our conclusions suggest caution in pushing universities to further expand their patent portfolios, following a trend common to all European countries for which data are available since the 1990s. The point was raised recently in Italy by the Committee for the Third Mission Evaluation, within the framework of the most recent national research assessment exercises (VQR 2011–2014). The Committee suggested not to link funding to indicators of patent ownership (as initially envisaged by the Ministry for University and Research) and for the first time produced several indicators based on all academic patents, not just the university-owned ones (see the Committee’s report: Committee for the Third Mission Evaluation of the Italian National Agency for University and Research Evaluation [CETM-ANVUR], 2017).
Relative to other directions taken by the literature on academic patenting, this has been characterized, among others, by a renewed interest in the effects of the professor’s privilege legislation on the level of academic patenting activity (Czarnitzki et al., 2016; Hvide and Jones, 2018), and more generally in the faculty-vs.-university IP assignment dilemma (Grimaldi et al., 2011). In this respect we find, in line with Lissoni et al. (2013), that the introduction and then the reform of the privilege in Italy has had no effect whatsoever on academic patenting. Within an historical perspective, this is not unexpected: whatever specific reform of IP legislation for universities one can conceive of, it will always have less impact than a major reform such as the introduction of university autonomy, as it occurred in Italy and more generally throughout Europe since the 1990s. Besides, as stressed by Halilem et al. (2017) for Canada, academic inventors decide whether to file patents through their universities or partner firms not merely in accordance with the specific IP legislation but also and especially on the basis of economic considerations (concerning their personal revenues as well as access to firms’ research funding and collaboration; see also Perkmann et al., 2013). In the Italian case, the university technology transfer managers we interviewed suggest that the contractual conditions for conducting research with partner firms are often drafted by the inventors’ university departments rather than central administrations, with the former having a preference for leaving IP in the firms’ hands in exchange for more funds and for keeping good terms with partner firms. This suggests a further level of complexity in IP assignment, one that escapes the simplistic contrast between individual scientists and university administration, as it may result from paying excessive attention to the professor’s privilege issue.
Footnotes
Notice that this terminology is not the same as that of less recent, US-oriented literature, which focused exclusively on university-owned academic patents, and referred to them alternatively as “academic patents” (as in Mowery and Ziedonis, 2007) or “university patents” (as in Henderson et al., 1998).
This new technique consisted in identifying academic patents, in particular those for which universities do not appear as applicants, by finding out the professional status of inventors, namely, to discover whether one or more among the latter had an academic status around the time of patent filing. This, in turn, required several data disambiguating inventors’ names and matching them to faculty’s names from university rosters or publications’ by-lines. For applications, see Callaert et al. (2013), Dornbusch et al. (2013), Iversen et al. (2007), Lawson (2013), Lawson and Sterzi (2014), Lissoni et al. (2009 and 2013), Martínez et al. (2013), Mejer (2012), Meyer (2003), Schoen and Buenstorf (2013), Thursby et al. (2009), and Sterzi (2013). For Germany, Czarnitzki et al. (2011 and 2012) classify patents as academic whenever the inventors’ name included reference to a professorial status. Notice that this line of research differs from the earlier, mostly US-centred one, largely centred on comparing university patents to corporate ones, while ignoring altogether the academic inventions being assigned to firms (Henderson et al., 1998; Mowery et al., 2001 and 2002; Rosell and Agrawal, 2009).
See references in previous footnote. See also Della Malva et al. (2013).
We borrow the term “IP university regulations” from Baldini et al. (2006) and Kochenkova et al. (2016), with reference to internal universities’ rules, and we use it contrast to “IP legislation,” which refers to national laws. But the terminology can vary: see for example the use of “institutional intellectual property rights policies” (IIPRP), by Halilem et al. (2017).
Article 65 in the new Italian Code of Industrial Property, introduced in 2005, maintains the privilege, but only for inventions stemming from research activities entirely funded by the university. The privilege is lost in presence of any research contract or convention for which the university has received external funding or co-funding, whether private or public.
See again references in footnote 2.
Czarnitzki et al. (2011) also find that academic patents receive on average more citations than all other patents, but that the difference appears to decline over time, along with the abolition of professor’s privilege
In most universities, including the Italian ones, the decisions concerning whether to file or sell or licence a university-owned patent is centralized at the top administration level. Departments, however, are entitled to sign research contracts and engage directly in research consortia with firms and other external organizations. These contracts and collaborations may include the cession of IP rights to the contracting firm, based on clauses which the central administration may not oppose, for legal or managerial reasons (avoid a clash with departments or influential members thereof).
To date, the Italian Government has run three evaluation exercises, at rather irregular intervals: VTR 2001–2003, VQR 2004–2010, and VQR 2011–2014 (Rebora and Turri, 2013; Geuna and Piolatto, 2016). All three evaluation exercises collected data on patents, albeit with different criteria. The consideration of university-owned patents in the formulation of ranking indicators was announced both for VQR 2004–2010 and VQR 2011–2014 (see, e.g., the ministerial decree DM 27/6/2015: http://attiministeriali.miur.it/anno-2015/giugno/dm-27062015.aspx), but ultimately not implemented.
Most public universities in Europe, including Italian ones, can hire their personnel only through tightly regulated job calls and have no freedom to increase their wages to attract the best. Regulations also exist that limit their freedom to pay for external consultancy services.
We do not consider all patents filed before 1997, as they are likely to include a large number of academic patents whose inventors were already retired in 2000, and thus escape our data collection effort. We do not consider patents first filed after 2009 to avoid right truncation bias of forward citations.
In a few cases, for patents with EPO priorities, we retrieved the information on applicants manually from Espacenet (https://worldwide.espacenet.com/). This is because, for such patents, PatStat automatically updates information and reports only the current applicants or assignees. So, whenever an EPO patent changes hands, information on its first applicant is lost.
Results are available upon request.
EPO patents can be renewed for up to 20 years since the priority date. A fee is due upon granting (for the granting to be effective) and then annually from the third year after granting. Hence, the theoretical value of Renewals runs from 1 (for patents just granted) to 20 (maximum patent length), minus the number of years lapsed since the priority date. In our database, Renewals range from 2 to 16, where 2 indicates a patent granted and never renewed, whereas the maximum at 16 indicates than no patent in our database has been yet renewed up to its expiry date (possibly owing to right-truncation problems in our observations).
For robustness check, we considered also all citations, irrespective of the timing. Results are very similar and available upon request.
Correspondingly, we interpret Share of Faculty in Team less than one as indicative of the likely presence of at least one industrial researcher in the inventor team, and the closer the variable is to zero, the larger the involvement of industry, which increases the probability of business ownership relative to university ownership. In principle, this is not necessarily the case, as also PhD students and post-docs are counted as nonpermanent faculty members, so that their presence in the inventors’ team would also return a lower-than-one value. However, these cases are a minority, as found by Lissoni (2010) and recently confirmed by ANVUR, the Italian Evaluation Agency for University and Research (CETM-ANVUR, 2017). Several informal interviews we undertook when collecting the data are in line with this evidence.
We owe the information on IP statutes to Baldini et al. (2010), whereas data on TTOs come both from official sources (Ministry of Education, University and Research—MIUR) and the NetVal survey of Italian TTOs (http://netval.it/servizi/survey/).
When in a patent document there are inventors affiliated to universities in two or more regions, we considered the maximum of this value.
Owing to right truncation, the maximum value of Renewals depends from the time span between the patent priority year and the most recent year covered by our patent database. The same problem affects, albeit much more moderately, Family Size, which may change over time as the decision of extending a patent abroad can be taken after the priority filing, namely, up to 12 months later for extensions through the Treaty of Paris route, and for up to 30 months for the Patent Cooperation Treaty (PCT) route. In addition, Family Size , as measured like we do with EPO’s DOCDB family, includes all the documentation produced by national patent offices after the publication, such as the publication of a granting decision or an opposition, which may occur further later in time.
As macro-classes in the first stage we simply distinguish patents in chemicals, pharmaceutics, and biotechnology (47.3% of granted patents in our sample) from all others (mostly from electronics and instruments). Broadly speaking, this reflects a distinction between “discrete” technologies (in which one or a few patents coincide with a new product or process, IP protection is effective, and licensing for a fee is quite common) and “complex” ones (in which the protection of a new product or process requires a combination of many patents as well as cross-licensing, owing to the fragmentation of IP (Cohen et al., 2000; Orsenigo and Sterzi, 2010). Notice however that, when applying PSM, we further control for technological classes at a finer level, as they enter through fixed effects in Equation 1. As for the matching method, our results are based on the “nearest neighbor with replacement” (Caliendo and Kopeinig 2008). We also conducted two additional robustness checks. In the first one, we imposed only the same priority year to each university-owned patent and its control. In the second one, we considered the seven macro-classes of the Schmoch’s (2008) classification and imposed the same one to each university-owned patent and its control. Results are unchanged.
For example, the early patents by a large and highly inventive university such as the Polytechnic of Milan will appear in the low-experience sample, before the total number of patents by the Polytechnic had reached the 14-patent threshold, whereas the more recent ones, filed after passing the threshold, will appear in the high-experience sample. On the contrary, patents by universities with an overall limited activity will only appear in the low-experience sample.
Notice that low-experience university-owned academic patents have more renewals and large families, on average, than high-experience ones, and that the same applies to the control patents for one and the other groups. This is owing to the composition effect discussed in the previous note. The high-experience sample contains a higher share of recent university-owned patents (and also of firm-owned controls, the matching criteria including the filing year). Owing to right-truncation problems, the more recent a patent, the lower the number of renewals we can observe. The same applies to the size of patent families, which include not only the first application at each national patent office but also the follow-up documentation. Observations of such documentation are also subject to right truncation. To put it differently, the table can be read only horizontally and not vertically, as patents in the high- and low-experience cells cannot be compared.
The composition effect does not affect our estimation of the university experience impact owing to the way in which we constructed our control sample. Indeed, every university-owned patent is matched to a firm-owned patent with the same priority year and each pair of matched patents has the same maximum achievable number of Renewals and Family Size. Therefore, (recent) high-experience university-owned academic patents are compared with (recent) high-experience firm-owned academic patents and, similarly, (older) low-experience university-owned academic patents are compared with (older) low-experience firm-owned academic patents. To put it differently, Table 6 can be read only horizontally and not vertically, as patents in the high- and low-experience cells cannot be compared.
Acknowledgments
Research for this article has been funded by Regional Council of Nouvelle Aquitaine (Chaire d’Accueil programme) and the French National Research Agency (“University Technology Transfer and its Optimization - UTTO” project; ANR-15-CE26-0005). Most data come from APE-INV, the Research Networking Programme on Academic Patenting in Europe funded by the European Science Foundation. Several colleagues kindly provided further data: Rosa Grimaldi, Riccardo Fini, and Maurizio Sobrero for universities' IP regulations and the TASTE database on academic spin-offs, and Massimo Colombo and Evila Piva for RITA data on academic spin-offs. Conversations with university technology transfer officers participating to the Committee for the Third Mission Evaluation of the Italian National Agency for University and Research Evaluation (CETM-ANVUR) proved particularly useful; in particular, Andrea Berti (University of Padova) dedicated us much of his time. Claire Sosso provided valuable research assistantship. Early versions of the article were presented to the FinKT project final conference (Rimini, April 2015), the EU-SPRI conference (Helsinki, June 2015), and the ICC conference “Beyond Technological Innovation and Diffusion” in Berkeley, CA (December 2016). All the discussants provided very useful comments. Any error remains ours.
References
CETM-ANVUR (
European Commission (