Moving beyond the paradox: Searching for the key factors ... - DIMETIC

Moving beyond the paradox: Searching for the key factors in research commercialization Evangelos Bourelos1, Mats Magnusson2 and Maureen McKelvey1 1University of Gothenburg (GU) Institute for Innovation and Entrepreneurship SE-405 30 Gothenburg, Sweden 2Royal Institute of Technology (KTH) Department of Machine Design SE-100 44 Stockholm, Sweden I. Introduction That economic growth is stimulated by innovation, research and development (R&D) and academic entrepreneurship is one key finding of innovation literature in recent years, and a basis for modern innovation policy (Fagerberg (2006)). The challenge for European policy in general – and Swedish policy in specific – has been the ‘paradox’ of high investment into R&D, but the low returns in growth (Edquist and McKelvey 1991). Dosi et al (2006), Ejermo (2005) among others have argued that the paradox argument is based upon an incorrect linear view from science to new products, and each author adds their specific recommendations and mechanisms for growth. This paper tackles the ‘paradox’ view in a different way, namely by examining whether European academics actually are quite so reluctant towards commercializing theirresearch as is commonly assumed in literature addressing the paradox. This question matters for combined theoretical reasons and public policy reasons. The theoretical reason contrasts the current focus upon the ‘paradox’ with work on the nature of problem-solving in technology. Early work by Rosenberg (1992), Pavitt (1998), and more recent work by Martin and Salter (2001), Lissoni et al. (2008) demonstrate that universities and university colleges interact in much more complex ways with industry than the usual focus of patents, technology transfer offices (TTOs) and start-up companies that are very much the current focus of attention. A key but provocative insight from Pavitt (1998) that we would like to bring back into the debate is that the university complements the industrial sector. This is diametrically opposed to the view that the university is a separate, substitute source of knowledge, which creates independent knowledge, which is economically useful and can be sold in the market through licensing, etc. We will further develop these ideas, in the context of ideas from different literature streams focusing on the issue of commercializing academic research. Developing these ideas has deep policy implications. If Pavitt’s (1998) insights are correct, then industrial and science policy must differentiate between different types of knowledge, with different rationales and mechanisms for interaction. This paper thus addresses two parallel but mutually reinforcing research questions: 1) Given the dominant negative view of academic entrepreneurship in Europe,, to what extent are Swedish researchers in engineering faculties involved in patenting and

start-up companies? And what are their attitudes, relative to competencies, in these two activities? 2) Theoretically, competing explanatory variables exist in the debates, and so what are the results, when related to each other? Can alternative theoretical perspectives be proposed to explain commercialization behavior and performance?

II. Literature Review 1. Academic entrepreneurship Academic entrepreneurship is an important element in modern economies as it can stimulate innovation and growth. Academic entrepreneurship is expressed by different outputs. Patents and start-up companies are usually measured as two direct means of transferring knowledge and technology from the universities to the market. However, the perceptions about how these processes actually work differ, as well as whether European systems lag behind American ones. Issues related to the supportive institutional and incentive structure for academic entrepreneurship are complex, because different levels of support structures will influence the outcomes in terms of commercialization. Three levels that can be clearly distinguished are the national institutional context, the university support structure, and the behavioral traits of the research group in which the individual is active (Magnusson et al., 2009). In the modern knowledge economy, universities should play a role in turning academic knowledge into economic wealth. Wherever market inefficiencies arise, universities can play an important substitute role (Grimaldi et al., 2006; Siegel et al., 2003). Therefore, the trend at most universities is that commercialization aspects of research are becoming an established part of their activities, and particularly in the biotech sector university business development emerges as a fundamental component of the value creation chain (Campbell, 2005). This is further reinforced by related factors such as the broader institutional framework, research funding pressures, institutional histories, culture of the university-department and peer influence (Etzkowitz, 2002; Kenney and Goe, 2004; Pelz and Andrews, 1976; Stuart and Ding, 2006). This suggests the need for university support structures, related to national policies on entrepreneurship. At the micro level, individuals matter, and they are often analyzed in relation to their research group/discipline and to career incentives. Technology transfer from universities is best facilitated by the active involvement of university inventors, and direct academic entrepreneurship is often seen as an effective means to facilitate technology transfer (Henrekson and Rosenberg, 2001; Slaughter and Leslie, 1997). Thornton (1999) points out two factors that influence a scientist’s decision to undertake commercialization activities, and the supply refers to individual’s characteristics and the demand to related agency mechanisms. Some literature argues that there is little yet known about the processes reshaping their career trajectories and pursuing entrepreneurial paths (Jain et al., 2009). Without going into social psychology literature, we do wish to suggest that individuals and their competencies and choices are important parts of commercialization.

Hence, academic entrepreneurship literature has two main levels of explanation, that of the support structure (as related to national policy and university decisions) and that of the individual (as related to competencies and incentives).

2. Key factors influencing commercialization Given the potential interplay of support structures and individuals, we wish to bring three different theoretical traditions together, which this far have usually been tested separately, coming from somewhat different fields. We wish to examine them together, in order to see if the results taken together are more coherent and if so, to possibly propose an alternative theoretical explanation. We are particularly interested in how ‘cognitive distance’ and cognitive aspects can help explain individual relationships and behavior, also over organizational boundaries. One set of literature suggests that there is a positive correlation between research performance and commercialization. The hypothesis is that researchers with numerous and well-cited publications, so called star scientists, are more likely to get involved in commercialization. These star scientists have relevant intellectual capital, and they are more likely to start firms and move into commercial involvement than are marginal scientists, as demonstrated within the specific research area of biotechnology, and later also in other areas (Zucker and Darby, 1996; Zucker et al. 1998). Similarly, other studies report a positive relationship between research quality and industrial interactions for university-industry interactions. Di Gregorio and Shane (2003), for example, found that intellectual eminence (in terms of the overall academic rating score of graduate schools published in Gourman Reports) was related to the number of spin-offs from universities. In contrast, work related to industry provides a more complex set of ideas about the value of science. Mansfield (1991 and 1998) provides hypotheses of why firms may be willing to work with the most scientifically advanced universities – as well as with the ones focused upon applied research like polytechnics. Thus, research performance can be of value to the individual star scientist through licensing and starting up companies as well as to the existing company in different ways. In this respect, engineering is particularly interesting to study further, as Mansfield (1991) and others make stronger predictions about the substitutability of basic research and applied research in these fields than in other disciplines. Nevertheless, there is no significant academic value, as a valuable advantage in academic position calls for example, in engineering from commercialization activities. It should here be noted that inputs and outputs may be linked. Research performance is often considered in terms of the ‘quality’ of the outputs, but research performance is also affected by inputs like grants won competitively. Merton (1959) argued that the Matthews effect lead to the most successful and known scientists obtaining more prestige and resources. Previous research performance is thus used as a signal to obtain more resources. Hence, based on this stream of literature, this paper will examine the possible correlations between grants, publications and commercialization. The expectation is that high performance in terms of papers and grants should be positively correlated in science-driven fields like biotechnology, but following Mansfield (1991), they could be either positively or negatively correlated in industrydriven fields like engineering.

Another set of literature instead stresses the role of social capital, residing in individuals’ networks. Hence, although it is difficult to outline it as a causal relationship, we also wish to examine the expected positive effects of individual researchers’ networks on commercialization. The network success hypothesis assumes a positive relationship between the networking activities of founders and the success of their start-ups. The rationale behind this hypothesis is the theory of socially embedded ties that allow entrepreneurs to get resources at a lower cost than they could be obtained in markets and to secure resources that would not be available in markets at all, e.g. reputation, customer contacts, etc. One prominent explanation for start-ups’ success has explicitly referred to network theory by investigating the personal networks of entrepreneurs and their effect on start-up performance (Birley, 1985; Aldrich, Rosen and Woodward, 1987; Johannisson, 1988). Another potential benefit of a central network position would be superior access to information and knowledge of various types, and then in particular complementary information flows, which taken together could increase the possibilities for both innovation opportunity recognition and the realization of such opportunities. One aspect of researcher networks is thus linkages to firms, either as a previous employee or through other type of interactions. Being a member of a board of a firm is probably a potential background which can push a research on commercializing. The relationship is again twofold. Usually, firms who use scientific advisory boards are those which are more dependent on scientific knowledge (Chok, 2009). Scientific advisory boards are also highly used in start-ups, especially in those where venture capitalists are involved (Isaacson, Mitchell, and Starr, 1994). Moreover a strong personal network indicates some typical attributes of the entrepreneur such as self-confidence, perseverance, resourcefulness, risk acceptance and achievement motivation (Hornaday, 1982). Successful entrepreneurs have been found to be more likely to have larger social networks that they can use to pursue and exploit opportunities (Gartner and Birley, 2002). Entrepreneurs who have been in business before are more successful in launching new businesses because of the existing networks (Vesper, 1980). Empirically, however, it is also interesting that empirical studies investigating networks’ effect on start-ups have rarely come up with any significant results (Witt 2004). A reason for this could be the difficult nature of quantifying and measuring personal networks, but the importance of networks is nevertheless not clear. This paper will study the effect of the researchers’ networks on commercializing, and it is important that the analysis of network relationships is fine-grained to distinguish different types of contacts. The third and final set of literature drawn upon in this study examines the support structures for commercialization. The support structure is mainly referred to the support that university provides toward commercialization. Etzkowitz (2004) argues that one important difference between Europe and the U.S. is that the entrepreneurial university emerged “bottom up” in the U.S. in contrast to Europe, where the involvement in commercialization by academia is a recent “top down” process. In additional to the ‘usual’ support structure of TTOs, our review also includes information about courses and training to students.

In a survey to TTOs at 57 UK universities, Lockett et al. (2003) found that the more successful universities now have clearer strategies towards the spinning out of companies, and also use “surrogate” entrepreneurs in this process. In addition, the more successful universities were found to possess a greater expertise and networks, and equity ownership was more widely distributed among the members of the spinout company. Hence, the result of this line of research stresses that the presence of support structures at a university, such as TTOs and incubators, can increase the chances of successful commercialization and thereby the output of inventions (Powers and McDougall, 2005). The development of the entrepreneurship field in the universities of a country is also important in order to grow a more positive commercialization environment. Brush et al. (2003) analyze the importance of doctoral studies in entrepreneurship in order to enrich and upgrade the field. Vesper and Cartner (1997) emphasize not only entrepreneurial courses in university but also the quality and the importance of the pedagogical teaching of entrepreneurship. Based upon this stream of literature, we will examine whether the academics used university support structures, including courses, and also the value of those structures for their attitude and action towards commercialization.

3. The Swedish context for commercialization of academic research In Sweden, it seems that it used to be less common for academics to start up a company, considering the trend of graduates to work in the public sector. During the 1970s two thirds of all academically trained people worked in the public sector. Although this share has decreased somewhat since then, it is still roughly 55 percent. (Henrekson and Rosenberg, 2001). Like the EU in general, Sweden has undergone many institutional reforms to try to stimulate research performance, as well as commercialization, to stimulate economic growth. In March 2000, the government presented two parallel and partly overlapping bills, dealing with the suggestions of Wigzell (1999) 1 and Flodström (1999) 2 respectively. Largely, the bills adopted academic peer review as the main mechanism behind public research funding, but went further than Wigzell by extending it far into the sphere of research based on sectoral needs (Eklund, 2005). Moreover, given the Swedish policy context, we expect granting also to be positively correlated with commercialization, assuming a reciprocal environment. To also support the causal effect of funding on commercialization, we should consider the more and more targeted innovation funding policies. During the period 1995-2001, the research funding system was reformed and after six years of heated debate, a reformed institutional structure for public research funding was finally established in Sweden on January 1, 2001, introducing VINNOVA, the Swedish Governmental Agency for Innovation Systems). VINNOVA is continuously providing funding to academia, with main targets to boost innovation activities, increase collaboration

1 2

Prop. 1999/2000:81, p. 14f. Ds 1999:68, p. 79ff.; Flodström (1999), p. 88f.

between universities, research institutes, companies and the public sector, and stimulate innovation based start-ups 3. Different researchers have put forward a rather negative view of academic entrepreneurship in Sweden. According to Hellström and Jacob (1999), Sweden has an ‘ivory tower’ culture embraced by researchers, which would suggest a lack of networks with companies. Johannisson (1988) points out the importance of personal networks as a key to entrepreneurial success and states that knowledge-based entrepreneurs are more engaged than other entrepreneurs in personal networking. He also attempts to quantify the concept of personal network which is usually assumed to carry subtle phenomena such as the entrepreneurial spirit or atmosphere of a region. An important result is that Swedish entrepreneurs who were about to launch a venture build considerably more socially oriented primary networks than those already in business (Johannisson, 1988). One might consider that a strong business network could probably be an additional motive for commercializing, especially in Sweden where it is said that there is a partial cultural prejudice on commercializing. Establishing university policies that promote the commercialization of research has been the mainstream Swedish efforts to assist academic entrepreneurship or other ways of technology transfer (Henrekson and Rosenberg, 2001). Swedish universities are said to be confronting several problems in the building up of such structures and policies. One view is that the Swedish government has invested lavishly in university research and enacted a set of policies to facilitate knowledge transfer, but it has failed to create a good incentive system for universities and academics to pursue the commercialization of ideas originating in academia (Goldfarb et al., 2003). Jacob et al. (2003) suggest that a considerable amount of tension still exists between those who view research as a public good and those who focus on the need to integrate university based knowledge production with the rest of the economy: Still, public policies in favor of strengthening the support structures in universities started already some decades ago. In 1975, a third objective was added to the agenda of universities, namely, to communicate to the surrounding society results emanating from university research and how they can be applied. Gradually this third objective came to be interpreted more broadly as collaboration between universities, on the one hand, and private industry and the public sector, on the other. In the new regulation of the universities, effective from 1998 (SOU 1998:128, pp. 153–154), this is spelled out explicitly. The universities are exhorted to be open to influences from the outside world, disseminate information about their teaching and research activities outside academia, and to facilitate for the surrounding society to gain access to relevant information about research results (Henrekson and Rosenberg, 2001). Throughout the 1980s and 1990s, STU and NUTEK collaborated with academic research groups and supported studies on innovation. The two most important groups for this were arguably the Sweden’s Technological Systems (STS) project led by Bo Carlsson and the network-oriented business economists headed by Håkan Håkansson at Uppsala University (Eklund, 2005).

3

VINNOVA, Research, Development and Innovation: Strategy Proposal for Sustainable Growth.

Establishing university policies that promote the commercialization of research has been the mainstream Swedish efforts to assist academic entrepreneurship or other ways of technology transfer (Henrekson and Rosenberg, 2001), and one of the main actors has been VINNOVA. Therefore, as far as money flows into universities in order to trigger commercialization, this study can also serve as a rough check for financing’s effectiveness. In summary, the three main variables we identify as most significant to test for academic entrepreneurship are: Research Performance, Personal Networks, and Support Structures. Personal networks are considered a key factor for entrepreneurship and in the academics’ case it is more specified in a case of potential nascent entrepreneurs. The second and third factors represent two special characteristics of academic entrepreneurs assumed to be of importance in the existing literature.

III. Empirical study 1. Methods used

As mentioned above, the three main independent variables used in the study are: Research Performance, Personal Networks and the Presence of Support Structures. In order to control for these variables we use a number of sub-variables for each one of the three categories. With this method, we do not only try to cover as much as possible the three factors, but we also expect to clarify which of the particular components of each factor is important for commercialization. A separate description of the division of the variables follows, so that the reader can see more clearly the grouping of variables we have done. With the Research Performance factor we refer to the number of publications and the grants each researcher receives. More specifically, publications consist of the number of papers, book chapters, peer reviewed articles and conference papers. The amount of grants can be separated as following; research funding by government, research funding by international governments, research funding by private foundations and research funding from business firms. A more complete measure of Research performance should cover a quality indicator as well, such as citations, but in this study we will use only self reported publication data. The fact that the literature, as well as intuition, suggests collinearity within the two explanatory sub-variables of Research Performance will be taken into account later in the econometric analysis. Due to the relatively small number of observations in the data and the large number of different variables, we try to reduce the number of variables which will be used in the regression models in order to have more robust results, considering also the nature of the data with most of the variables to contain qualitative data. The first step is to try to cluster the variables. Therefore, we try to cut down the variables of Research performance into two groups, namely Publications and Total Grants. The possibility

to merge the variables is dependent on the reliability statistics we perform, including the Cronbach alphas and the correlation matrices.

The Personal Network variable includes the time spent by a single researcher in interacting with the academic and business environment, more specifically time spent with other university researchers, researchers in industrial firms, entrepreneurs, venture capitalists and with managers in industrial firms. As before, we try to merge the variables, naming the five mentioned types of interaction NetworkTime1. The Personal Network variable also includes what kind of cooperation with firms that a researcher has been engaged over for the five last years. The different kind of cooperations referred to here are: Contract research for a company, Working in a joint research project, Publishing joint research articles, Consultant for university’s benefit, Consultant for the researcher’s own benefit, Meetings with firms, Supervision of PhDs co-financed by firms, Training of company employees, Helping students to be placed in firms, and being a Member of a board of a firm. Altogether, these variables are aggregated as the variable NetworkTime2. The presence of support structures includes the following variables: Following courses on entrepreneurship, use of incubator services, use of TTO services, opinion towards spinoffs, opinions towards patents, opinion towards commercialization. The first three denote which means a researcher has used and the next three refer to his or her personal opinions. The last three variables are used as proxy variables to count for the general entrepreneurial environment as an element of the broad support structure. The three first variables constitute the variable SupportStructure1 and the latter three the variable SupportStructure2. In order to check which variables can be finally clustered in the regression models, we apply a number reliability tests. We use as a rule of thumb for our clustering a value of alpha greater than 0.6 together with positive correlations higher than 0.2.As we see in tables A.1.a, A.1.b and A.1.c in the appendix A, the four variables in the scale publications are positively correlated to each other and alpha is big enough to allow us the clustering. Based on the heuristics, the variables referring to the total grants do not overpass the clustering thresholds as we put them before, because as we see in tables A.2.a and A.2.b the value of alpha is 0.5 and the correlations between the variables are not all higher than 0.2. Nevertheless, we extend the tolerance for this case and include it in the regression as we consider it an important component of the model. On the other hand, NetworkTime2 is eligible to be used in the models because of the high correlations and alpha value (tables A.4.a and A.4.b). As is also seen in Tables A.3.b and A.3.c, NetworkTime1 has a value of alpha close to 0.5, but it increases to more than 0.6 if the variable “timewithotheruniversity researchers” is deleted. Therefore, we use this variable separated from the others and cluster the rest of them. Finally, SupportStructures1 consist only of three variables which are highly correlated each other, but they give a value of alpha of only about 0.55 (table A.5.b). Nevertheless, we will also cluster these three variables with this value of alpha because of the small amount of variables and the fact that they are correlated each other. SupportStructures2 is a proper cluster, but we do not finally use it in the regressions because there is a non negligible lack of observations to all of its variables provoking problems in the models.

The dependent variable of the different analyses is the commercialization intensity, which counts the total amount of companies founded and patents sold out by each researcher. Commercialization intensity is a more complete measure than a binary, as it gives information not only about which researchers have commercialized but also about the degree to which they have done it. Apart from the main dependent and independent variables presented, we will also employ some other additional variables as regressors in order to be able to explain more of the variation of the dependent variable. The additional variables that we control for are primarily; gender, year of birth, university, discipline, the highest degree obtained, if the researcher has been employed by a firm, how frequently commercialization is discussed, if the researcher has considered commercialization and finally his or her opinion about patenting. The most important of these additional variables is the one indicating if a researcher has been employed by a firm. A strong connection with a firm, such as an employment relationship is assumed to be highly correlated with commercialization. We also assume that considering commercializing and a positive opinion on patents may have a positive effect. Another result which would not surprise us is a gender effect, taking into account that in the field there is already a gender’s uneven composition in favor of men. (77% of the researchers were men). The dataset contains a large number of variables, with most of them to be count or binary variables, as they express qualitative information. The binary variable “indexcomm” is not used as a dependent variable because being a binary variable it would require logistic regression. The biggest disadvantage of the logistic regression, in this case, is that there are no adequate observations for a reliable logistic regression. Researchers are not in complete agreement as to how big sample sizes need to be in order to obtain stable estimates. Long (1997) suggests that sample sizes of less than 100 should be avoided and that 500 observations should be adequate for almost any situation. However, this leaves a relatively large gap between 100 and 500 4. Here we have 207 observations used for the regression of 32 variables of interest. Moreover, for 80 percent of the sample the response variable was equal to 0. Therefore we cannot make very safe conclusions from the logistic regression. In order to get some more reliable estimates, we ran a regression on the count variable “commercialization intensity”, which counts the total amount of companies founded and patents sold out by each researcher, and therefore is more informative than the variable “indexcomm”. As we can see in Figure 1. in the appendix A, the data is strongly skewed to the right, so clearly OLS regression would be inappropriate. Count data often follows a Poisson distribution, so some type of Poisson analysis might be appropriate. As known from statistical theory, in a Poisson distribution the mean and variance are the same. As we can see in table A.6 in the appendix A the variance of “commercialization intensity” is nearly four times larger than the mean. Therefore, there is a clear sign of overdispersion, i.e.- a greater variance than might be expected in a Poisson distribution, indicating that the Poisson distribution might be inappropriate. For this reason, we employ a negative binomial regression in order to count for the over-dispersion problems. In the negative binomial regression, the likelihood ratio test at the bottom of the analysis is a test of the overdispersion 4

UCLA Academic Technology Services

parameter alpha. When the overdispersion parameter is zero, the negative binomial distribution is equivalent to a Poisson distribution. When alpha is significantly different from zero, the Poisson distribution is not the most appropriate. The negative binomial regression in the end seems to be the most appropriate for this dataset. After applying it in the clustered variables, we also apply it in a model with all the variables of interest separate, resulting in a regression with 17 more variables. For comparison reasons we also apply the Tobit regression in the clustered data, which is an alternative for similar data. Tobit is useful for this kind of limited dependent variable which is roughly continuous over strictly positive values but is zero for a non-trivial fraction of the population. We do not expect to see huge differences between the Tobit model and the negative binomial distribution, as the data seems to fit both models.

2. Data collection The data used in this study was collected using a web-based survey. The survey was designed after a thorough review of literature, and the aim was to elicit data that would increase our understanding of Swedish academic researchers’ involvement in the commercialization of university research. Commercialization was here defined as consisting of two main activities, the generation of start-up firms, and the generation of patents that are licensed or sold out. In addition, the survey results should help us identify differences between researchers who have commercialized their findings and those researchers who have not been involved in such activities, as related to the three elements in the selection environment earlier mentioned. Survey questions therefore focused on whether individuals were positive or negative towards different commercialization activities, the individuals’ attitudes towards commercialization, the researchers’ knowledge about entrepreneurship and commercialization, their possible use of university support structures, and communication with industry and broader institutions. These questions were included, as they represented possible explanations for differences between commercializers and non-commercializers that we identified in the undertaken literature review. The survey was sent to a little more than 1200 researchers employed at Swedish universities and university colleges, representing a number of personnel categories which can be assumed to engage at least to some extent in research. Six research areas of engineering and natural science were chosen, namely: Fluid Mechanics, Inorganic Chemistry, Wood Technology, Computer Science, Biotechnology and Automatic Control. We chose these particular fields because these are all areas where at least one of the universities hosts a national competence center. VINNOVA, the Swedish Agency for Innovation Systems, set up these VINNEX Centers, with the explicit aims ‘to become an academic, multidisciplinary Centre of Excellence by actively involving a number of companies in joint research’ and ‘to promote the implementation of new technology and to strengthen the technical competence in Swedish industry’. All other universities or university colleges with centers and departments in the same research area were then identified and included in the sample. Hence, while only one of these universities have a specific VINN Excellence Centers in any given area, the research

field per se has been identified as one of interest to companies and to Swedish industry. Taking these VINNEX centers and the other departments in the same research fields together resulted in a total population from different universities. Note that each university may have more than one relevant center or department. This then resulted in a survey sent to all researchers in the 50 departments in the six research areas, distributed according to the following list. For a complete list of departments included, see Appendix B. 1) Fluid Mechanics (5 departments at 4 different universities) 2) Inorganic Chemistry (6 departments at 5 different universities) 3) Wood Technology (7 departments at 6 different universities) 4) Computer Science (10 departments at 7 different universities) 5) Automatic Control (11 departments at 6 different universities) 6) Biotechnology (11 departments at 6 different universities) Before sending out the survey, we initially contacted the heads of the departments (or research leaders) for approval to conduct it. The survey was performed using a webbased questionnaire, with the initial query and reminders sent out during a period of three months. The data in the created database was analyzed using SPSS and STATA software. The population is quite large, to be inclusive of these six research areas, and hence the total population was composed of 1219 academic researchers. The response rate was 24.2 percent, leaving us with usable responses from 295 researchers. The response rate is thus relatively low as compared to surveys on other topics conducted in Sweden, but about the same rate as surveys on university-industry relations conducted in other countries. Still, some reservations should be made for the comparably low response rate, which most certainly was influenced by the large number of questions in the web-based questionnaire. We therefore analyzed for representativeness of the sample, as compared to the total population receiving the questionnaire. To avoid bias of over- and under-representation of specific research areas, we also checked the response rate by research area, e.g. aggregated for all universities within a specific area (Magnusson et al., 2009).

3. Descriptive Statistics In our sample, 19 percent of the researcher reported that they are involved in a commercialization activity. 7 percent reported only 1 patent or 1 start up. The average reported number of publications was approximately 19 publications per researcher. Of these, about 7 were peer reviewed articles, 5 conference papers and 6 were other papers. On average, each researcher had received a grant of more than 100000 SEK twice. Regarding their meetings, the researchers spent most of their time with other university researchers, thereafter in falling order with researchers from industrial firms, managers, entrepreneurs and venture capitalists. Regarding their activities, the researchers spent a large part of their time on joint research, which is in accordance with the time spent with other university researchers. The order of time spent from higher to smaller is the following: time in firms, time in supervising PhDs, time spent on contract research, time spent on helping students to be placed in a firm, time used

for consultancy, time spent on training company employees, and time as a member of a board. Table A.7 presents means and standard deviations for the variables used in the study.

4. Model In the first model, equation (1), we apply a negative binomial regression model with the main interest dependent variables clustered. The goodness of fit in the negative binomial regression is calculated based on the Pearson residuals. In the second model, equation (2) is again a negative binomial regression with the use of all the specific variables and without any compressed variables. In detail, the two different models are as follows:

log(E(commercializationintensity)) = α + β1 publications_hgt + β 2 Grants_hgt + β3 SupportStructures1_hgt + β 4 NetworkTime1_hgt + β 5 NetworkTime2_hgt + β 6 gender + β 7 yearofbirth + β8 university2 + β9 simplegroupnum + β10 highestdegree2 + β11 employedinprivatefirm + β12 FreqDiscCommercCollegu + β13 ConsidCommerciyReser + β14 opiniononpatent + β15 timeotheruniversityresearcher (1)

log(E(commercializationintensity)) = α + β1 Npapers + β 2 Nbookchapters + β3 Npeerreviewarticles + β 4 Nconferencepapers + β 5 researchfundingbygovernmentc + β 6 ResearchFundingIntGovern + β 7 ResearchFundingPrivFound + β8 ResearchFundingBYBusinessFirms + β9 timeresearchersinindustrialfirms + β10 Timeentrepreneurse + timeinvestorsventurecapitalistsb + β11 β12 timemanagersinindustrialfirms + β13 usedTTO + β14 Usedincubators + β15 Usedcourses + β16 Timecontractresearch + β17 Timejointresearch + β18 Timeconsultancy + β19 Timemeetingswithfirms + β 20 Timememberofaboardofafirm + β 21 Timesupervisedphds + β 22 timetrainedcompanyemployees + β 23 gender3 + β 24 Yearofbirth + β 25 university2 + β 26 Simplegroupnum + β 27 highestdegree2 + β 28 Employedinprivatefirm + FreqDiscCommercCollegu + β 29 β30 ConsidCommerciyReser + β31 Opiniononpatent + β32 Timeotheruniversityresearcher (2)

5. Results and analysis This section presents the results and analysis of the performed study. In Table A.8, we can see the results from the negative binomial regression, when we use the clusters in model 1. Table A.9 displays the results from the negative regression for the extended model with all the variables used in model 2. We also run regressions with Tobit models but the results were in all relevant aspects similar and are not reported. 5.1 Negative Binomial regression - clustered model Table A.8 exhibits the coefficients and standard errors from the regression results for equation 1. The dependent variable used is “Commintensity” in a negative binomial regression. Significant variables are “NetworkTime2”, “employedinprivatefirm” and “FreqDiscCommercCollegu”. Having been employed in a private firm is, as expected, a positive correlation with commercialization. According to the negative binomial regression used here, researchers who are, or have been employed in a private firm have 71% higher commercialization output than other researchers, controlling for the other variables. The negative coefficient of the “NetworkTime2” can be explained by the fact that “NetworkTime2” is a clustered variable and therefore the different components inside the cluster have different effects. The negative coefficient might also be an indication of positive effects of some of the other time related variables inside the “NetworkTime1” cluster, as the total available time of a researcher is limited, and the different time variables consequently substitute each other. Therefore, a clearer picture of the effects of the network components follows in the results of the next model, in which clusters were not used. The same stands for the other dependent variables of main interest because, as seen above, the overall effect of the clusters did not appear to be significant. At the bottom of Table A.8 also the likelihood ratio test is displayed.. In this case, the associated chi-squared value is 38.97 with one degree of freedom and this once again underscores that the Poisson distribution is not the most appropriate one. 5.2 Negative regression - unclustered model Table A.9 exhibits the regression results and standard errors on equation 2. The equation includes the same dependent variable as before, but using a total of 32 independent variables. The input of more information in the model, after the separation of the variables, results in specifying the factors which affect commercialization. Apart from that, research performance and support structures which did not appear significant, as estimated in aggregate forms, now become important through the specific variables of peer-reviewed articles, used courses and the use of incubators. These are all positively correlated with commercialization, with used courses standing out as the variable with the strongest interdependence. We can also distinguish some different and contradicting effects of personal networks through a number of important variables. The most important variables from the personal network, with positive correlation to commercialization, are the time spent with entrepreneurs and the time spent on consultancy. Moreover, time spent together with managers in industrial firms is also a positive and significant variable. On the other

hand, the time spent as a member of a board of a firm, time spent in meetings with firms, time spent with researchers in industrial firms, and time spent with venture capitalists all have negative effects. The contradictory effects of the time spending variables may be a result of the substitution effect, as time has to be shared across the different activities and there may be an opportunity cost for time spent in different activities. We have to mention that the question was phrased as “What percentage of your working time do you spend on each of these activities” and therefore the time has to be shared by definition. In order to check if there is a particular pair substitution effect, we also calculate the Pearson correlation coefficients in Tables A.10 and A.11, for the two nexus of time variables. In Table A.10 where we have the first nexus, in correspondence with the questionnaire, we see that time with entrepreneurs is negatively correlated with time with other university researchers. The same stands for the time with managers in industrial firms. In Table A.11, for nexus 2, there is no pair with negative correlation. Nevertheless, the overall substitution effect of time is a reasonable explanation for the negative coefficients on some time variables, as mentioned above. Regarding the control variables, employment by a firm is also in this model significant and strongly positive. Considering commercialization of research is also positive as expected. The gender factor, which appears positive, is not very useful for further justification because the sample is already strongly dominated by men. The likelihood ratio test is again presented at the bottom of Table A.9, and also in this case alpha is significantly different from zero and the negative binomial distribution is thus appropriate. Summarizing the empirical findings, certain factors at an individual level appear to have a significant influence on the commercialization. In particular these characteristics are the use of courses in entrepreneurship and related fields, and mobility between industry and academia.. These positive correlations closely related to the individual’s commercialization capability gives a clear innovation policy message. In order to make commercialization thrive, a certain level of individual commercialization knowledge and skills appears to be necessary. The main contribution of entrepreneurial training and industrial experience is likely to enable researchers to see opportunities and inspire them towards entrepreneurship. The argument that the relation between opportunity recognition and training can be in the opposite direction is of course reasonable, but considering that we in our analysis control for attitudes with proxy control variables, the positive effect of training through courses indicates that training is the cause and not the effect in that case. An explanation of this importance on individual’s capability could be that the system is lacking the influence of the researcher as a very first step before being able to benefit from TTOs and other support structures such as incubators. That could be also an explanation of TTOs’ non-significance in the model. If the very first step of university commercialization, that is to provide researchers with the necessary motivation and skills, is not taken care of then TTOs cannot have a valuable effect by themselves. Based on this observation, it can be argued that incentives from the university should more explicitly attend to the researchers’ motivation, knowledge and skills. For example, recognizing the academic value of industrial work experience could be an incentive for boundary-spanning mobility, thereby possibly increasing the researchers ‘commercialization efforts In terms of concrete actions, this would imply that, the academic community, and in particular university management need to include

entrepreneurial activities as a valuable component in the university researcher career process.

6. Discussion and conclusions Sweden has been criticized for having a low ‘entrepreneurial spirit’ (Goldfarb et al., 2003) and a lack of a supportive institutional and incentive structure (Henrekson and Rosenberg 2001). A key notion of the more recent version of the ‘Swedish paradox’ and ‘EU paradox’ is that the countries are good at science and bad at commercialization (see critical reviews in Dosi et al., 2006; Ejermo, 2005). Using an extensive survey of engineering researchers, this paper presents strikingly different results. This leads us to a different interpretation of the university research commercialization situation in Sweden and also a different interpretation of the theoretically interesting variables and the relevant policy responses. In contrast to expectations based upon the paradox argument, we find that the Swedish researchers surveyed both engage in and are positive towards commercializing the results of their research. The argued ‘lack of entrepreneurial spirit’ suggests that the results expected for Sweden should be that researchers do not engage in commercialization activities and also should have a negative attitude towards it. Our empirical study shows that 11.5% of the included researchers have started a company and/or taken a patent and that 76% are positive or very positive towards commercialization. Hence, as compared to international figures, Swedish university researchers in these six fields of engineering research do start up new companies and also patent – and an overwhelming high percentage are positive to commercialization. 5 Apparently, the notion that there is a lack of entrepreneurial spirit among Swedish researchers need to be questioned, at least if we refer to researchers in the engineering disciplines investigated in this study.

This paper has analyzed the relative importance of three theoretical variables for explaining the commercialization activities of Swedish engineering researchers, namely research performance, personal networks, and support structures. We find that the picture that emerges is rather complex, which suggests that we first and foremost need to re-examine the explanatory variables underlying commercialization. Research performance is expressed by publishing and grants. Stephan et al. (2007) find that publishing has positive and significant effect on patenting. In a subset of scientists in the biomedical area in the US, no sign for substitution effect is found either, (Geuna and Nesta, 2006). Geuna and Nesta propose that this effect could be different depending on the age of the researcher. They therefore suggest testing the hypothesis that older researchers have the ability to patent and publish at the same time, because of the accumulated human capital. Young researchers, on the contrary, 5

Lissoni et al 2009 have recalculated university-invented patents (as opposed to university-owned patents) and find a similar percentage of academic patents in France, Italy, Sweden as in the USA. The reason previous research about Sweden has found few patents is that they focus upon patents held by the university, which in Sweden is about 5% of total academic patents.

are more likely to show substitution effect. Patenting, on the other hand, is considered not to substitute publishing but rather to trigger it(Owen-Smith (2003), Breschi et al. (2007); According to our empirical observations, research performance is clearly positively related to commercialization, at least when this performance is measured in terms of peer-reviewed articles. . This implies that some aspects of the ‘star scientist’ hypothesis hold in Sweden, in that the most active researchers are those who commercialize their research. In contrast to what Mansfield (1998) argues, we do not find a substitution effect between publishing and commercialization, something Mansfield (1991) suggests could be visible in engineering fields. However, if one instead examines grants, the star scientist hypothesis is not completely supported. In other words, the engineering researchers still publish, regardless of their funding, and those who publish more also tend to commercialize more. (Note that all categories of personnel are included in the sample, so it is not just a question of senior star scientists). Considering the direct effects of funding, two previous studies by Gulbrandsen and Smeby (2005) in Norway and one by Ranga (2003) focusing on the Katholik Universiteit van Leuven (KUL) in Belgium found positive statistical associations between industrial funding of university research and university patenting activity. This correlation could also reflect the star scientist hypothesis, indicating that the talented academic researchers would attract external research support whether or not they generated patents (Geuna and Nesta, 2006). In our analysis, a similar effect did not appear, as the industrial funding variable did not appear to be significant. We instead had the same results for all the funding variables. Nevertheless, we have to mention here that it is unclear if there is an indirect effect of funding. If the Mathew effect applies in this case then there is a correlation of the two variables of the research performance, publications and funding. The Mathew effect suggests that; research groups that are successful in finding external funding for their research have a higher probability of producing publishable research, which improves their probability of getting funds in the future (Geuna, 1999). In the performed study, funding could have an indirect positive effect on commercializing in our sample through the positive effect in publishing. Statistical support for this hypothesis comes from the fact that publications and grants are clearly and positive correlated and from the fact that if we run the model without the variables of funding, then the variable of peer reviewed articles loses its significance. Furthermore, funding could be affected by networks, given that it is likely that funding also changes network configurations. (Gulbrandsen, Smeby 2005) Summarizing the above, we can conclude that although there is no direct effect of funding on commercializing, there appears to be positive indirect effect. Based on the performed study, it is seen that personal networks are also significantly linked to commercialization, but also that different networks components have opposite effects. Time spent on consultancy for firms as well as time spent with entrepreneurs have positive effects on commercialization, just like time spent with managers in industrial firms . An interesting finding is that meetings with firms, time with venture capitalists and time with researchers in industrial firms, and time spent as a member of a board all are negatively correlated with commercialization, but this can likely be interpreted as a substitution effect against time for consultancy, time with entrepreneurs, and time with managers in industrial firms. The important conclusion here is that spending time with entrepreneurs and as a consultant,

compared to the other components of personal networks, has a positive effect on commercialization. It should be stressed that the overall effect is not clear, and it is important to underline that the performed study does not reveal the causal relationships between networks and commercialization, but only correlations. In that sense, it appears logical that there is a relationship between time spent with entrepreneurs and involvement in commercialization activities. The effect of consultancy is in this sense less straightforward, but could rather be explained in terms of providing insights regarding the market use of performed research. Support structure are positively linked to commercialization in terms of courses taken and the use of incubators, which are in turn linked to a clearly positive average opinion at the individual level towards both patents and start-up companies. The most interesting part of this is that we thus interpret that the individual researchers benefit the most from courses and hence development of individual competencies O’Shea et al. (2007) have suggested that there is a contribution of the supporting organizational mechanisms, such as TLO, in the entrepreneurial success of MIT. In our case, TTO was not a significant variable, rather supporting a skeptical view of TTOs effectiveness. Nelsen, (1998) as well as Charles and Conway (2001) claim that the costs of TTOs are not counterbalanced by the patent or copyright license incomes. Contrasting this,, Azagra Caro and Llerena (2003) find some empirical evidence to support the view that the laboratories of more prestigious groups (in terms of institutional recognition) tended to patent more. Of course, the latter could also be a result of better incubators in the laboratories in question. An explanation of the insignificance of TTO in our model could also be the missing first step of incentives to the individual, as we proposed already in the analysis. Summarizing the effects of the three investigated sets of variables (and research traditions), this paper shows that commercialization, measured as patents and start-up companies, is positively correlated to research performance in terms of peer reviewed articles , to personal networks linked to companies, to courses taken in the fields of entrepreneurship and commercialization, and to the use of incubators. Other variables with positive significance are previous experience from working in a company and a positive view of the individual researcher on commercialization. Theoretically, we can propose that university researchers who have industrial experience and/or use their personal networks to consult for companies have a low cognitive distance to companies (Broström and McKelvey, 2009), and they therefore may have better competencies to commercialize and likely position themselves in a network. This may imply that having contacts with companies through personal networks and personal experience working in a company are two factors of great importance for the ability to patent and start companies. Thinking logically, these results seem logical. However, we feel that current policy – with the exception of a few initiatives aiming at stimulating labor mobility – has often neglected this aspect. The implications of these findings for policy would be to focus upon activities that further mobility between academia and industry and give researchers increased possibilities to gain work experience from business firms – and possibly put less focus upon issues more narrowly focused on the direct realization of patents, TTOs and start-up companies. These activities may very well play a role in the overall commercialization complex, but unless more researchers perceive commercialization opportunities and are willing to pursue them, the support structures will not have the input

Our interpretation would be that experience with the world of business helps the researcher – not only in commercialization but possibly also to do more ‘science’, as measured by papers. We feel that the insights of Rosenberg (1992) and Pavitt (1998) thus should be re-introduced into the debate, because they argue that company interaction helps the researcher at the research institute or university to specify and define new sets of research problems. This result can be placed in relation to other recent findings about life sciences in medical technologies and in food in Sweden (Laage-Hellman et al., 2009). These detailed case studies also reveal that researchers who publish are active in a broader range of commercialization activities, and that the large public policy initiatives have a positive impact upon publications. Another more direct effect of publishing on commercialization can be the legitimacy and publicity they can provide to companies, especially that they may help generating value already early on in projects, as Laage-Hellman et al. (2009) suggest in a VINNOVA project on life science. Having here the evidence that publishing does not substitute commercializing as well as the evidence of the individual ‘s importance, we can go one step forward, combining the results in order to suggest innovation policies. And this combination suggests that we should focus in changing the attitudes and replace the dilemma of publishing vs. commercializing. We should focus on the identity of the actions that both outputs need and on their complementarity. The change on the attitudes ground can be succeeded by modification of the incentives. Thus, in order to gain radical changes on the ground level, policy should focus on the individual incentives. Taken together, the results in this paper and the case study observations in the mentioned work by Laage-Hellman et al. (2009) suggest that Swedish researchers whom are successful in publications also interact with industry and commercialize – leaving us to strongly question whether there is a gap between research and commercialization in the ‘paradox’, when analyzed at the level of the individual researcher. This line of research is in our view interesting to develop further, given that one limitation of this paper is that we only examine commercialization in terms of patents and start-up companies. Future research on the later performance of those companies should also be relevant, given that research suggests that these academic spin-offs grow slower than others – but when they do grow, they are vital for economic growth. Our limited indicators for commercialization – and the positive results that we find – are also interesting for another reason. A line of research has provided very strong empirical evidence for the many ways in which universities and their researchers contribute to economic growth (Martin and Salter, 2001). Future research could investigate the many ways in which these university researchers interact with companies – but even barring that richer future data, it appears plausible that more contact (and experience) with companies helps the researcher to act as academic entrepreneurs and possibly also to be better researchers in terms of their scientific performance, being better informed about relevant questions to target in their research studies.

Note that in this particular paper, we focus upon how the individuals view the university support structure. An imitation of the assumed American model has been developed in Sweden since the late 1990s– with universities pushing their IPR, with more extensive use of TTOs, and with the development of ‘knowledge and innovation environments’ at universities. Our results are quite interesting in this context. For public policy initiatives, the university support structures are discussed in terms of TTOs , but the TTOs variable does not show up as significant. In other words, despite the huge emphasis in Sweden upon the need for the university to hold IPR and to develop its own support structure, where the developments in the 1990s and 2000s have been to strengthen them to make up for perceived deficiencies in the national institutional context, the results show that they actually do not impact commercialization in this sample. This may be because they still have not reached the maturity level required to improve commercialization, and thus need to develop competencies – but it may also be because Swedish university-industry interaction occurs in other ways than normally picked up by indicators or normally focused upon by policy-makers. We call for an examination of the other types of academic entrepreneurship identified here – as well as a realization that the ‘culture’ of academic entrepreneurship is in fact very lively, and more complex than usually realized. This has far-going implications for public policy, to focus upon labor mobility, work experience, networks and meeting points. Finally, this means that university and public policy efforts to commercialize are not wrong – but their limitations and where they work poorly must be acknowledged. If public policy does not take these specific and very different structures into account as to interactions in the knowledge economy, then European public policy runs the risk of destroying features of knowledge development and diffusion that help structural transformation in the national innovation systems. In other words, with an incorrect analysis of the problem, European public policy makers may not only develop inappropriate public policy responses – but they may even destroy less tangible but functioning strengths of the current knowledge economies.

References Aldrich, H., Rosen, B., & Woodward, W. (1987). The impact of social networks on business foundings and profit: A longitudinal study. Frontiers of entrepreneurship research, 7, 154-168. Azagra Caro, J. M., Llerena, P. (2003). Types of contractual funding and university patents: from analysis to a case study. [Conference Paper]. Paper Prepared for Knowledge and Economic and Social Change: NewChallenges to Innovation Studies Conference, Manchester, CRIC. Baldini, N., Grimaldi, R., & Sobrero, M. (2006). Institutional changes and the commercialization of academic knowledge: A study of Italian universities' patenting activities between 1965 and 2002. Research Policy, 35(4), 518-532. Birley, S. (1985). The role of networks in the entrepreneurial process. Journal of Business Venturing, 1(2), 107-117. Breschi, S., Lissoni, F., & Montobbio, F. (2007). The scientific productivity of academic inventors: new evidence from italian data. Economics of Innovation and New Technology, 16, 101-118. Broström, A., & McKelvey, M. (2009). Universities and public research institutes as collaboration partners. Work in progress. Brush, C., Duhaime, I., Gartner, W., Stewart, A., Katz, J., Hitt, M., et al. (2003). Doctoral education in the field of entrepreneurship. Journal of Management, 29(3), 309. Campbell, A. (2005). The evolving concept of value add in university commercialisation. Journal of Commercial Biotechnology, 11(4), 337-345. Charles, D., Conway, C., (2001). Higher Education–Business Interaction Survey, A Report to the UK HE Funding Bodies (HEFCE, SHEFC, HEFCW and DEL) and the Office of Science and Technology. Chok, J. (2009). Regulatory dependence and Scientific Advisory Boards. Research Policy, 38(5), 710-725. Di Gregorio, D., & Shane, S. (2003). Why do some universities generate more startups than others? Research Policy, 32(2), 209-227. Dosi, G., Llerena, P., & Labini, M. S. (2006). The relationships between science, technologies and their industrial exploitation: An illustration through the myths and realities of the so-called [`]European Paradox'. Research Policy, 35(10), 1450-1464. Edquist, C., & McKelvey, M. (1991). The diffusion of new product technologies and productivity growth in Swedish industry', Consortium on Competitiveness and Cooperation (CCC) Working Paper No. 91-15, 1992. Center for Research in Management, Berkeley, CA: University of California. Ejermo, O. (2005). Country Report: Efforts and Performance of R&D in Sweden. In Efforts and Performance of R&D in the Nordic Countries, ed. by S. Jespersen Efforts and Performance of R&D in the Nordic Countries, ed. by S. Jespersen[Copenhagen: Ministry of Finance.]. Eklund, M. (2005). Reception of the Innovation System Concept in Sweden. Paper presented at the Ekonomisk-historiska mötet. Eklund, M. (2007). Adoption of the Innovation System Concept in Sweden. Universitetsbiblioteket, Uppsala.

Etzkowitz, H. (2002). Incubation of incubators: innovation as a triple helix of university-industry-government networks. Science and Public Policy, 29(2), 115-128. Etzkowitz, H. (2004). The evolution of the entrepreneurial university. International Journal of Technology and Globalisation, 1(1), 64-77. Fagerberg, J., Mowery, D., & Nelson, R. (2006). The Oxford handbook of innovation: Oxford University Press, USA. Gartner, W., & Birley, S. (2002). Introduction to the special issue on qualitative methods in entrepreneurship research. Journal of Business Venturing, 17(5), 387-395. Geuna, A. (1999). The economics of knowledge production : funding and the structure of university research. Cheltenham: Elgar. Geuna, A., & Nesta, L. J. J. (2006). University patenting and its effects on academic research: The emerging European evidence. Research Policy, 35(6), 790-807. Goldfarb, B., & Henrekson, M. (2003). Bottom-up versus top-down policies towards the commercialization of university intellectual property. Research Policy, 32(4), 639-658. Gulbrandsen, M., & Smeby, J.-C. (2005). Industry funding and university professors' research performance. Research Policy, 34(6), 932-950. HellstrOm, T., & Jacob, M. (1999). Evaluating and Managing the Performance of University-Industry Partnerships: From Central Rule to Dynamic Research Networks. Evaluation, 5(3), 330. Henrekson, M., & Rosenberg, N. (2001). Designing efficient institutions for sciencebased entrepreneurship: lesson from the US and Sweden. The Journal of Technology Transfer, 26(3), 207-231. Hornaday, J. (1982). Research about living entrepreneurs. Encyclopedia of entrepreneurship, 21-22. Isaacson, R., Mitchell, R., & Starr, L. (1994). Getting the most from a scientific advisory board. Research Technology Management, 37, 33-33. Jacob, M., Lundqvist, M., & Hellsmark, H. (2003). Entrepreneurial transformations in the Swedish University system: the case of Chalmers University of Technology. Research Policy, 32(9), 1555-1568. Jain, S., Gerald, G., & Maltarich, M. (2009). Academics or entrepreneurs? Investigating role identity modification of university scientists involved in commercialization activity. Research Policy. Jens Laage-Hellman, M. M. M. J. (2009). Analysis of Chain-linked Effects of Public Policy-Effects on research and industry in Swedish life sciences within innovative food and medical technology. VINNOVA - Swedish Governmental Agency for Innovation Systems/Verket för Innovationssystem, Vinnova Analysis VA 2009:20. Johannisson, B. (1988). Business formation-a network approach. Scandinavian Journal of Management, 4(3/4), 83-99. Kenney, M., & Richard Goe, W. (2004). The role of social embeddedness in professorial entrepreneurship: a comparison of electrical engineering and computer science at UC Berkeley and Stanford. Research Policy, 33(5), 691707. Lissoni, F., Llerena, P., McKelvey, M., & Sanditov, B. (2008). Academic patenting in Europe: new evidence from the KEINS database. Research Evaluation, 17(2), 87-102.

Lockett, A., Wright, M., & Franklin, S. (2003). Technology transfer and universities' spin-out strategies. Small Business Economics, 20(2), 185-200. Long, J. S. (1997). Regression models for categorical and limited dependent variables: Thousand Oaks, CA, US: Sage Publications, Inc. Magnusson, M., McKelvey, M., & Versiglioni, M. (2009). 9. The forgotten individuals: attitudes and skills in academic commercialization in Sweden. Learning to Compete in European Universities: From Social Institution to Knowledge Business, 219. Mansfield, E. (1991). Academic research and industrial innovation. Research Policy, 20(1), 1-12. Mansfield, E. (1998). Academic research and industrial innovation: An update of empirical findings. Research Policy, 26(7-8), 773-776. Merton, R. (1959). Notes on problem-finding in sociology. Sociology today: Problems and prospects, 1. Nelson, L. (1998). Essays on science and society: The Rise of Intellectual Property Protection in The American University. Science, 279(5356), 1460-1461. O’Shea, R., Chugh, H., & Allen, T. (2008). Determinants and consequences of university spinoff activity: a conceptual framework. The Journal of Technology Transfer, 33(6), 653-666. Owen-Smith, J., & Powell, W. W. (2003). The expanding role of university patenting in the life sciences: assessing the importance of experience and connectivity. Research Policy, 32(9), 1695-1711. Pavitt, K. (1998). The social shaping of the national science base. Research Policy, 27(8), 793-805. Pelz, D. C., & Andrews, F. M. (1976). Scientists in organizations. Mich.: Institute for Social Research, University of Michigan, Rev. ed. Ann Arbor. Powers, J., & McDougall, P. (2005). University start-up formation and technology licensing with firms that go public: a resource-based view of academic entrepreneurship. Journal of Business Venturing, 20(3), 291-311. Ranga, L., Debackere, K., & Tunzelmann, N. (2003). Entrepreneurial universities and the dynamics of academic knowledge production: A case study of basic vs. applied research in Belgium. Scientometrics, 58(2), 301-320. Rosenberg, N. (1992). Scientific instrumentation and university research. Research Policy, 21(4), 381-390. Salter, A. J., & Martin, B. R. (2001). The economic benefits of publicly funded basic research: a critical review. Research Policy, 30(3), 509-532. Siegel, D. S., Waldman, D., & Link, A. (2003). Assessing the impact of organizational practices on the relative productivity of university technology transfer offices: an exploratory study. Research Policy, 32(1), 27-48. Slaughter, S., & Leslie, L. (1997). Academic Capitalism: Politics, Policies, and the Entrepreneurial University: The Johns Hopkins University Press, 2715 North Charles Street, Baltimore, MD 21218-4319 ($39.95). Stephan, P. E., Gurmu, S., Sumell, A. J., & Black, G. (2007). Who's patenting in the university? Evidence from the survey of doctorate recipients. Economics of Innovation and New Technology, 16(2), 71 - 99. Stuart, T., & Ding, W. (2006). When Do Scientists Become Entrepreneurs? The Social Structural Antecedents of Commercial Activity in the Academic Life Sciences 1. American Journal of Sociology, 112(1), 97-144. Thornton, P. (1999). The sociology of entrepreneurship. Annual Review of Sociology, 25(1), 19-46.

Vesper, K., & Gartner, W. (1997). Measuring progress in entrepreneurship education. Journal of Business Venturing, 12(5), 403-421. Vesper Karl, H. (1980). New Venture Strategies: Englewood Cliffs, NJ: Prentice-Hall. Witt, P. (2004). Entrepreneurs. Networks and the Success of Start-ups’, Entrepreneurship and Regional Development, 16(5), 391-413. Zucker, L., & Darby, M. (1996). Star scientists and institutional transformation: Patterns of invention and innovation in the formation of the biotechnology industry. Proceedings of the National Academy of Sciences, 93(23), 12709. Zucker, L., Darby, M., & Brewer, M. (1998). Intellectual human capital and the birth of US biotechnology enterprises. The American Economic Review, 88(1), 290306. Government BillsProp. 1999/2000:81 Forskning för framtiden – en ny organisation för forskningsfinansiering.Government Reports Ds 1999:68 Att finansiera forskning och utveckling: rapport från Arbetsgruppen för fortsatt beredning av myndighetsstruktur för forskningsfinansiering. SOU 1998:128, Forskningspolitik. Slutbetänkande av Kommittén för översyn av den svenska forskningspolitiken (Forskning 2000). Stockholm: Fritzes.

APPENDICES

Appendix A.

Scale: IndexPublications

Table A.1.a: Inter-Item Correlation Matrix Published Number of

Book chapters

Articles

conference

papers published

published

published

papers

Number of papers published

1,000

,379

,574

,379

Book chapters published

,379

1,000

,450

,250

Articles published

,574

,450

1,000

,572

Published conference papers

,379

,250

,572

1,000

Table A.1.b: Reliability Statistics Cronbach's Alpha Based on Cronbach's

Standardized

Alpha

Items ,695

N of Items ,754

4

Table A.1.c: Item-Total Statistics Cronbach's Squared Multiple

Alpha if Item

Correlation

Deleted

Number of papers published

,352

,570

Book chapters published

,225

,749

Articles published

,519

,477

Published conference papers

,332

,590

Scale: TotalGrants

Table A.2.a: Inter-Item Correlation Matrix How many times How many times had received

from int.

more than

Council/agencies

100000 sek from

, probably

government

international it

from Private

by bussiness

council/agencie

means

foundations

firms

How many times had

How many times How many times

1,000

,200

,154

,349

,200

1,000

,216

,250

,154

,216

1,000

,142

,349

,250

,142

1,000

received more than 100000 sek from government council/agencie How many times from int. Council/agencies, probably international it means How many times from Private foundations How many times by bussiness firms

Table A.2.b: Reliability Statistics Cronbach's Alpha Based on Cronbach's

Standardized

Alpha

Items ,507

N of Items ,528

4

Table A.2.c: Item-Total Statistics Cronbach's

How many times had

Squared Multiple

Alpha if Item

Correlation

Deleted

,143

,428

,106

,440

,063

,496

received more than 100000 sek from government council/agencie How many times from int. Council/agencies, probably international it means How many times from Private foundations

Table A.2.c: Item-Total Statistics Cronbach's Squared Multiple

Alpha if Item

Correlation

Deleted

How many times had

,143

,428

,106

,440

,063

,496

,159

,357

received more than 100000 sek from government council/agencie How many times from int. Council/agencies, probably international it means How many times from Private foundations How many times by bussiness firms

Scale: NetworkTime

Table A.3.a: Inter-Item Correlation Matrix Time spent with other

Time spent with other

researchers,

Time spent with

1=none, 7=large

researchers in

time with

part

industrial firms

entrepreneurs

1,000

,125

-,044

,125

1,000

,257

-,044

,257

1,000

,089

,174

,624

-,015

,414

,458

researchers, 1=none, 7=large part Time spent with researchers in industrial firms time with entrepreneurs time with venture capitalists time with managers

’

Table A.3.b: Reliability Statistics Cronbach's Alpha Based on Cronbach's

Standardized

Alpha

Items ,500

N of Items ,614

5

Table A.3.c: Item-Total Statistics Cronbach's Squared Multiple

Alpha if Item

Correlation

Deleted

Time spent with other

,047

,662

,196

,369

time with entrepreneurs

,472

,401

time with venture capitalists

,406

,421

time with managers

,309

,365

researchers, 1=none, 7=large part Time spent with researchers in industrial firms

Scale: NetworkTime2 Table A.4.a: Inter-Item Correlation Matrix Time contract research 0-7 ,

Time contract research 0-7 ,

1=never, 7=very

Time Joint

much, 0=do not

research 0-7 as

Time

Time meetings

know

before

consultancy 0-7

with firms 0-7

1,000

,423

,452

,376

,423

1,000

,418

,520

Time consultancy 0-7

,452

,418

1,000

,439

Time meetings with firms 0-7

,376

,520

,439

1,000

Time supervised phds co

,224

,400

,205

,449

1=never, 7=very much, 0=do not know Time Joint research 0-7 as before

financed with a firm, 0-7

Time trained company

,279

,372

,392

,438

,229

,383

,351

,414

,276

,152

,359

,279

employess 0-7 Time helped students to be placed in firms, 0-7 Time spent with member of a firm 0-7

Table A.4.b: Reliability Statistics Cronbach's Alpha Based on Cronbach's

Standardized

Alpha

Items ,822

N of Items ,831

8

Table A.4.c: Item-Total Statistics Cronbach's

Time contract research 0-7 ,

Squared Multiple

Alpha if Item

Correlation

Deleted

,297

,813

,411

,798

Time consultancy 0-7

,363

,803

Time meetings with firms 0-7

,418

,789

Time supervised phds co

,340

,809

,502

,792

,501

,792

,385

,814

1=never, 7=very much, 0=do not know Time Joint research 0-7 as before

financed with a firm, 0-7 Time trained company employess 0-7 Time helped students to be placed in firms, 0-7 Time spent with member of a firm 0-7

Scale: SupportStructures1

Table A.5.a: Inter-Item Correlation Matrix binary 0, 1, If there is a transfer technology office

binary if

at your university

binary if used

participated to

have y

incubator

courses

binary 0, 1, If there is a

1,000

,288

,312

binary if used incubator

,288

1,000

,288

binary if participated to

,312

,288

1,000

transfer technology office at your university have y

courses

Table A.5.b: Reliability Statistics Cronbach's Alpha Based on Cronbach's

Standardized

Alpha

Items ,547

N of Items ,558

3

Table A.5.c:Item-Total Statistics Cronbach's

binary 0, 1, If there is a

Squared Multiple

Alpha if Item

Correlation

Deleted

,140

,426

binary if used incubator

,127

,476

binary if participated to

,140

,426

transfer technology office at your university have y

courses

Table A.6 • . summarize commercializationintensity, detail • counts the total amount of companies founded and • patents sold out by each resear • ------------------------------------------------------------• Percentiles Smallest • 1% 0 0 • 5% 0 0 • 10% 0 0 Obs 270 • 25% 0 0 Sum of Wgt. 270 • 50% 0 Mean .2259259 • Largest Std. Dev. .9271183 • 75% 0 3 • 90% 1 4 Variance .8595484 • 95% 1 8 Skewness 7.221455 • 99% 4 10 Kurtosis 66.54855

Table A.7: Means and Standard Deviations Variable Commercializationintensity

Mean

Std. Dev. 0,226

0,927

6,830

20,441

0,521

1,257

0,037

0,214

0,694

1,433

3,080

2,843

5,936

13,307

0,675

1,869

7,159

17,589

4,759

11,673

1,041

1,360

0,369

0,758

publications_hgt Grants_hgt SupportStructures1_hgt NetworkTime1_hgt NetworkTime2_hgt Npapers Nbookchapters Npeerreviewarticles Nconferencepapers Researchfundingbygovernmentc ResearchFundingIntGovern

ResearchFundingPrivFound 0,220

0,625

0,342

0,874

4,573 2,847 1,556 1,292 1,892

1,901 1,523 0,931 0,721 1,235

0,138

0,346

0,076

0,265

0,147

0,355

2,276

1,790

3,369

2,065

2,020

1,469

2,924

1,715

2,488

2,032

1,941

1,401

2,129

1,620

1,383

0,999

0,769

0,422

1968,512

10,830

5,658

2,924

3,563

1,617

1,678

0,826

0,278

0,449

1,376 4,417

1,039 1,789

ResearchFundingBYBusinessFirms timeotheru~r Timemanagersinindustrialfirms Timeentrepreneurse Timeinvestorsventurecapitalistsb Timeresearchersinindustrialfirms UsedTTO Usedincubators Usedcourses Timecontractresearch Timejointresearch Timeconsultancy Timemeetingswithfirms Timesupervisedphds Timetrainedcompanyemployees Timehelpedstudentstob Timememberofaboardofafirm gender3 Yearofbirth university2 Simplegroupnum highestdegree2 Employedinprivatefirm FreqDiscCommercCollegu ConsidCommerciyReser

Opiniononpatent 4,809

1,559

Table A.8: Regression results, clustered Dependent Variable: commercializationintensity (1) commercializationintensity

VARIABLES publications_hgt Grants_hgt SupportStructures1_hgt NetworkTime1_hgt NetworkTime2_hgt gender3 Yearofbirth university2 Simplegroupnum highestdegree2 Employedinprivatefirm FreqDiscCommercCollegu ConsidCommerciyReser Opiniononpatent Timeotheruniversityresearcher Constant

Observations Likelihood-ratio test of alpha=0:

0.0136 (0.0437) -14.98 (1,188) -16.71 (11,561) -0.245 (0.290) -0.274* (0.145) 1.714 (1.190) -0.00786 (0.0258) -0.0848 (0.0946) -0.185 (0.182) 0.372 (0.368) 1.186** (0.540) 0.537* (0.280) 0.298 (0.195) 0.223 (0.172) -0.234 (0.146) 10.73 (51.33)

206 chibar2(01) = 38.97 Prob>=chibar2 = .000 Standard errors in parentheses *** p