Revisiting Technological Centrality in University-Industry Interactions: A Study of Firms Academic Patents

Transcription

1 Revisiting Technological Centrality in University-Industry Interactions: A Study of Firms Academic Patents Maureen McKelvey, Evangelos Bourelos and Daniel Ljungberg* Institute for Innovations and Entrepreneurship, School of Business, Economics and Law, University of Gothenburg, Sweden * Abstract This paper examines whether and how large firms engage academic inventors in their invention process to diversify their technology base ( exploration ). Here we provide an analysis of firm patents, focusing on two sub-questions related to this purpose: i) To what extent do large firms engage academics as inventors in their core technologies?; and ii) Does academic engagement affect (large) firms technology base over time, and if so how? We contribute empirically by analysing firm-owned academic patents based upon new systematic evidence from the KEINS/APE-INV database on Swedish academic patents. Our preliminary descriptive findings suggests that large firms mostly engage academic inventors in their invention process in noncore technology fields, but that these fields over time take a more central position in the firms technology profile. Keywords: Academic patenting; University-industry interaction; Technological diversification; Local search 1

2 1. Introduction Although technology and innovation are often postulated as key drivers of economic growth, we know surprisingly little about how and why large firms rely upon academic researchers and universities to identify and exploit technological opportunities. A huge stream of literature on university-industry relations has focused upon the perspective of the university, and how knowledge diffuses outward. This includes explanations about why and how academics may succeed or fail or be disinterested in commercialization and in the wider concept of academic engagement with industry (Perkmann et al., 2012). Few papers, however, start from the perspective of the firm, to ask how academics contribute to industrial inventions. This paper therefore contributes through an exploratory study. Our contribution is by examining the technological centrality of inventions resulting from universityindustry relations, relative to the firms technological profile over time. Literature from evolutionary economics and the knowledge-based theory of the firm provides insights to developing the concept of technological opportunities. Research and development (R&D) was especially seen as a main source of new opportunities. Scherer (1965) put forth the argument that technological opportunities help explain the differential propensity of firms in different sectors to patent and innovate. He argues that differences in technological opportunity e.g. differences in technical investment possibilities unrelated to the mere volume of sales and typically opened up by the broad advance of knowledge are a major factor responsible for interindustry differences in inventive output (Scherer, 1965: 1121). This type of literature often tries to explain major differences amongst industries, and the role of technological knowledge in stimulating growth. Breschi et al. ( ) argue that observed sectoral patterns of innovative activities are related to the nature of the relevant technological regimes/ / defined by the specific combination of technological opportunities, appropriability of innovations, cumulativeness of technical advance and the properties of the knowledge base underlying firms innovative activities. This concept thus places the firms own investment into research and development activities in the idea that their probability of finding an invention and later innovation is dependent not only on the firm per se, but upon the likelihood of discovering new knowledge and the overall rate of change in this type of knowledge. 2

3 In literature on search for industrial inventions, the concept of search is a key aspect of finding and developing those opportunities. In this literature, it is argued that firms inventive activities are closely related to and bounded by their existing knowledge base (e.g. Nelson and Winter, 1982; Fleming, 2001; Rosenkopf and Nerkar, 2001; Nickerson and Zenger, 2004). This is often referred to as local search, meaning the behavior of firms to innovate by primarily relying upon their own previous knowledge and competencies. To put it differently, firms incrementally search close to or in their available knowledge base for a solution to a given problem when engaging in inventive activities (e.g. Fleming, 2001; Nelson and Winter 2001; Nickerson and Zenger, 2004). This is seen as local search. Firms can in principal move beyond local search by spanning their technological boundaries or organizational boundaries. This means that they can access new knowledge residing in distant technological domains (i.e. technological diversification) and/or by turning to external sources of knowledge (Rosenkopf and Nerkar, 2001). Studies have suggested the importance of combining knowledge across technologies as well as of accessing external knowledge (e.g. Fleming 2001; Rosenkopf and Nerkar, 2001; Rosenkopf and Ameida, 2003). Finding new sources of ideas externally for technological opportunities is thought to be key in order to innovate and remain competitive. Previous studies have mainly focused on other firms as external sources of knowledge, through e.g. strategic alliances and mobility of inventors (e.g. Rosenkopf and Almeida, 2003). Different measures of technological distance and cognitive distance have been developed to measure how close, or far, the knowledge or organizations are in terms of understanding science, technology, and markets (Nooteboom, 2009). This paper explores another potentially important source of external knowledge, academic researchers and universities. Universities and academics have been shown to be relatively important external sources for innovation (e.g. Mansfield, 1991; 1998; Cohen et al., 2002), even though they usually rank lower than market-related contacts like suppliers and customers. However, there are hardly any studies, which directly examine the issue of whether and how academic researchers and universities 3

4 contribute to firms exploration of new areas, which leads to technology diversification as compared to the existing technological profile of the firm. One exception is a study by Santoro and Chakrabarti (2002). They conclude that industrial firms do not typically use university relationships to help strengthen and build core competencies, but rather that large firms focus upon non-core technologies when interacting with universities and academic scientists, while small firms focus upon core technologies. This study uses survey-based methods and they investigated collaboration within university-industry research centres. Hence, their conceptualization and results suggest that large firms do and ought to engage academic researchers and universities to develop technological areas that are far away from the core technologies in the firms technological profiles. In this paper, we examine whether and how firms do indeed engage academics in their invention process to diversify their technology base, or we could say, use them for exploration. Large firms are more likely and more able to be technologically diversified and this should be increasingly so over time (Granstrand et al., 1997). They are also the types of firms more likely to interact with universities (e.g. Fontana et al., 2006). This paper focuses on large firms (defined as more than 250 employees following OECD), given the interesting question about large firms and the area of technological opportunities, following the results of Santoro and Chakrabarti (2002). This would mean that firms engage academic researchers and universities mainly for exploration of highly novel areas, and this seems reasonable, given the large potential through science and basic research. The empirical analysis is based upon patent data, comparing firm-owned academic patents with the total set of patents held by the same firm but not involving academic inventors. 1 This is done by drawing on new systematic evidence from the APE-INV database on Swedish academic patents (see Lissoni et al., 2006 for an overview of a previous version of this database), using two full datasets from and This paper presents preliminary findings, which are primarily a descriptive analysis, focusing on two related questions. From the framing in the literature, we can see that 1 Academic patent denotes a patent with one or several inventors affiliated to a university, regardless of assignee. 4

5 by engaging academics or interacting with universities in the invention process, the firm is by definition engaged in an organizational boundary spanning activity. This should help the firm overcome the tendency toward local search, and find new knowledge and opportunities for innovation and competitive advantage. But are these technological opportunities central to the existing firm profile? Or do large firms engage academic inventors for overcoming the barriers of local search? In that case, we would expect that university-industry interactions should be in areas that are quite different from the firms core profiles, and in that sense, the organizational boundary spanning activities ought to be employed for diversifying their technology base ( exploration ). 2. Literature review and research questions Researchers in neo-schumpeterian and business innovation literature have proposed that business inventions are largely the outcome of recombination of existing knowledge or technologies (e.g. Fleming, 2001; Nelson and Winter, 1982). Simplified, the invention processes can be described as: i) identifying and choosing a (valuable) problem, and ii) searching for solutions to this problem. 2 From this perspective of the knowledge-based theory of the firm, knowledge generation is promoted through searching for and choosing problems with potentially valuable solutions (Nickerson and Zenger, 2004). 3 Therefore, invention can be defined as resulting from the search for potential solutions to an identified problem, where a new solution is equivalent to a new combination, or a reconfiguration of existing combinations (cf. Fleming, 2001). In this view, invention is either a new combination of technological components 4 (Nelson and Winter, 1982) or an invention is a reconfiguration of existing combinations of technological assets (Henderson and Clark, 1990). Thus, inventive 2 This is a stylized depiction, concerning deliberate search, but problems are in reality identified also in other ways, e.g. through serendipity. It can also take the opposite route: a solution in search of a problem. 3 This could be about problems, and their solutions, on many different levels in the firm: organizational, marketing, R&D etc. We are here only concerned with problems regarding inventions (R&D). 4 A component represents any fundamental part of knowledge or physical artefact or matter used to construct an invention. 5

6 activities can be characterized as a search process for new or reconfigured combinations of knowledge and technologies (Fleming and Sorenson, 2001). This means that existing skills and knowledge, as well as access to those held by others, are useful to make new combinations. From the firm s perspective, they need to make a choice about the problems to pursue, such as how to develop an innovation or develop a technology. Doing so is based upon a combination of factors, especially the expected value of possible solutions, the ability to appropriate this value and the cost of searching for a solution (Nickerson and Zenger, 2004). Due to cognitive limitations of the inventor(s), and the fundamental uncertainty inherent in inventive processes, this choice is closely related to the firms prior knowledge and experiences. Accordingly, this literature suggests that firm inventions tend to be about local search for solutions (Fleming, 2001; Nelson and Winter, 1982). This means that it is more likely that the firm will incrementally search in the available knowledge base for a solution to a given problem. Thus, firms inventive activities are local in the sense of being closely related to the firm s existing knowledge base. They have a strong tendency to look for technology and innovations in areas they already know. Their technology base, or technological profile can be divided into core technologies and other types. Firms are, in this way, bounded by their available knowledge bases (and competencies), and when they do invent, their prior knowledge and experience guide the identification of problems and the search for solutions. However, while firms inventive activities tend to be local, their search processes are not necessarily limited within the boundaries of the firm. Firms, especially larger firms, tend to be technologically diversified, and increasingly broaden and deepen their technology base over time (Granstrand et al., 1997). Moreover, this means the firm needs to span the organizational boundaries as well as technological boundaries of the firm (Nelson and Winter, 1982; Nickerson and Zenger, 2004; Rosenkopf and Nerkar, 2001). Indeed, at times the firm can explore and exploit technological opportunities externally, because they may be required to search for problems and/or their solutions outside the firm as well as outside the boundaries of the existing technology base. 6

7 In fact, studies have examined firms partnerships with other firms from this perspective, and the results suggested the importance of combining knowledge across technologies as well as of accessing external knowledge (e.g. Fleming 2001; Rosenkopf and Nerkar, 2001; Rosenkopf and Ameida, 2003). In order to innovate and remain competitive, firms need to leverage their existing knowledge base by accessing new (external) knowledge to overcome their tendency towards local search in the innovation process (Fleming, 2001; Nickerson and Zenger, 2004; Rosenkopf and Nerkar, 2001). Or put differently, the firm must explore new possibilities rather than just exploit old certainties (March, 1991). As introduced above, Rosenkopf and Nerkar (2001) suggest that firms in principal move beyond local search by spanning their technological and/or organizational boundaries. What is interesting here is the idea that the firm can access new knowledge residing in distant technological domains (i.e. technological diversification), which may occur by turning to external sources of knowledge. These authors provide a typology of search over technological and organizational boundaries. 5 Local search is the internal search process inside the firm where the inventor is (incrementally) searching in the available knowledge base drawing on similar technologies as those present in-house, whereas radical search spans both technological and organizational boundaries. Internal boundary-spanning search is searching technologies within the firm that are relatively unrelated to the knowledge base, while external boundary-spanning search is the other way around. The question is how to understand firm s interactions with academic researchers and universities, from this conceptualization of industrial invention. Involving academics in industrial inventions can be interpreted as spanning the organizational, and potentially also the technological, boundaries, and thereby academic engagement in industry can help overcome the tendency toward local search. 6 If we interpret within 5 Technological boundaries are less straightforward than organizational, but have to do with the (degree of) relatedness of a technology with the current knowledge base of the firm. Thus, technological boundary spanning is a matter of degree, meaning that searching technologies with a relatively low, but not non-existing, relatedness to the firm s current knowledge base can be seen as (partly) spanning the technological boundaries. 6 This paper examines firm-owned patents with academics as (co-)inventors, assuming that the underlying inventions result from some degree of collaboration between academics and firms. Thus, while firms academic patents may result from arms-length collaboration, such as firms commissioning academics to conduct contract research on their behalf, they do not result from a market for 7

8 the above taxonomy of Rosenkopf and Nerkar (2001), we can say that academic inventors are involved in firms invention process for external boundary spanning search. However, should these university-industry relations lead to industrial inventions that are closely related to the existing technology base of the firm? Or should they overcome local search problems and lead to radical search? The question we therefore focus on in this paper is whether (large) firms can and do engage academic inventors for overcoming also technological localization, i.e. for diversifying their technology base, and whether and how this engagement affect firms technology base over time, Academics can be conceptualized to play different roles, depending on the firms needs for searching for solutions and/or problems across organizational boundaries. Santoro and Chakrabarti (2002) studied collaboration within university-industry research centres, concluding that industrial firms do not typically use university relationships to help strengthen and build core competencies. Their conclusion is that large firms, in contrast to smaller firms, focus on interacting with universities and academic scientists to build competencies in non-core technological areas. This finding suggests that large firms interact with universities mainly for diversifying their technology base (i.e. radical search). Moreover, it leads to a prediction over time. Over time, this collaboration will also build up the firms competencies, and thereby lead to a transformation of the technology base with previously distant technologies taking a more central position in the firm s technological profile. The finding of Santoro and Chakrabarti (2002) might however be related to the fact that university-industry research centers provide a setting which might be more conducive for collaborating on fundamental research rather than to develop specific technologies or solve specific technical problems (Feller et al., 2002). At the same time, do other studies suggest that firms mainly engage academics to aid in already technology where academics independently conduct research and firms later acquire the invention or patent. Rather, the inventive activities, i.e. the search for solutions and/or problems, are conducted jointly between firms and academics, with firms involving academics in their inventive activities and the resulting patents are directly assigned to the firms. 8

9 on-going R&D projects (Cohen et al., 2002; Perkmann and Walsh, 2009). Seeing that firms in-house R&D activities tend to be local, i.e. building and extending incrementally upon the existing technology base, we could predict that academic inventors are engaged mainly in firms core technologies, and thus do represent external boundary spanning but not leading to technology diversification. One empirical papers suggests that the large firms use academic researchers to build up existing core competencies and only occasionally to diversify and transform the technology base (Ljungberg and McKelvey, 2012). 7 To summarize, the existing literature proposes two different answers to the question of whether firms engage academics in their invention process to overcome their tendency towards local search by diversifying their technology base. As such, the literature provides little guidance for drawing any clear conjectures related to the questions raised here, and therefore this paper can mainly be seen as an explorative study. This leads us to two research questions: First, to what extent do large firms engage academics as inventors in their core technologies? Put differently, we examine the relation between academic engagement and the technology centrality of the resulting inventions. Second, does academic engagement affect (large) firms technology base over time, and if so how? 3. Method 3.1 Data and sample We developed a dataset of firm-owned patents where at least one inventor is employed at a Swedish university, starting from a set of academic patents sampled from the KEINS/APE-INV database. This is a comprehensive database on European academic patents, containing European Patent Office (EPO) patent applications that have been matched by name and surname with complete lists of university researchers of all ranks (from assistant to full professor). In this way, patents can be identified as 7 This is quite close to survey-based studies on university-industry interaction, which tends to find that firms perceive the greatest benefits of academic engagement to be for assisting in technical problemsolving and product development (e.g. Bishop et al., 2011; Lee, 2000). 9

10 academic also when the university is not the applicant. (For a detailed account of the KEINS database and its construction, see Lissoni et al., 2006.) From the KEINS/APE-INV database, we extracted all firm-owned Swedish 8 academic patent applications with priority years Thus, in line with our purpose, we did not include patent applications owned by individuals or non-firm organizations, such as universities. The complete list of Swedish academics used for identifying academic inventors in the APE-INV database was compiled by collecting personnel lists from Swedish universities in 2004 and 2011, meaning that only those academics active in those years were identified. Thus, the present sample lacks patents invented by academics that retired or changed profession before We reviewed each identified firm assignee, by exploring annual reports and similar sources, in order to identify the large firm assignees in the original sample. We here follow the OECD definition and consider firms with more than 250 employees as large. The database was cleaned from patent applications assigned to firms, which did not fulfil this criterion. PATSTAT was used to identify all non-academic patent applications owned by the same large firms identified as assignees of the sampled academic patents, as well as to collect all other patent data such as patent class and priority year. Thus, we sampled the entire population of patents assigned to a set of specific firms, rather than employing a random or matching sampling, which is the most common approach in the literature (cf. e.g. Henderson et al., 1998; Mowery et al., 2002). For comparison purposes, we only included (non-academic) patent application having at least one Swedish based inventor, since we only had access to Swedish academic patents. Delimiting the analysis geographically to Sweden, to the extent possible, lowers the risk of including patents having academic inventors affiliated to non-swedish 8 That is, all patents with at least one inventor affiliated to a Swedish university. 9 We included patent applications regardless of whether they have been granted or not, since we want to analyze academic inventions in general. 10

11 universities. After cleaning the data, the sample consists of patents, including 480 academic patents, owned by 16 firms. 3.2 Variables To conceptualize the relatedness of academic patents to the firms technology base (i.e. technology diversification), we follow Granstrand et al. (1997) and Patel and Pavitt (1997) that propose the concept of technological profiles. This is a four-field taxonomy, based on two dimensions of technological classes of firms patents (see Figure 1). The concept of technological profiles highlights differences between technological fields in terms of the firms commitment to the technology (patent share) and the importance of that technology for the firms competitive advantage (revealed technological advantage). Figure 1. Technological profiles. Adapted from Patel and Pavitt (1997) Following the taxonomy of Granstrand et al. (1997), core competencies consist of technological fields where the firm commits a high share of its resources (i.e. high patent share), and commands a strong technological advantage as compared to the overall population. Background competencies are fields where the firm has a relatively high commitment, but this does not translate into a strong advantage compared to other firms. Marginal competencies are fields with a low commitment, 11

12 i.e. currently relatively unrelated to the firm s technology base, and low competitive advantage. In marginal fields, firms do not (yet) command a strong commitment or advantage, but they may provide potentially strong opportunities for the future if the firm is able to build up competencies and a competitive advantage. Niche competencies are areas where the firm has a strong position, but does not dedicate a high share of its inventive activities and resources. From this conceptualization, we distinguish between firms core and non-core technologies by constructing the technological profiles of the firms. The technological profiles are defined according to two dimensions of patent classes, namely Patent share and Revealed technology advantage (RTA) (see Granstrand et al., 1997). Accordingly, we construct two measures: Patent share (PS): the share of firms patents in the technological class of the focal patent. This measure tells us whether the focal patent is in a technological class relatively closer to or further away from the firm s overall patent portfolio. Revealed technology advantage (RTA): the share of the firm in total patenting in the technological class of the focal patent, divided by the firm s aggregate share in all classes. In this way, we get a measure that tells us how important a firm is for a technological field relative to the overall population of firms. 10 The general interpretation of the RTA is that if it takes value over 1, then that firm has an "advantage" in that class over the overall population. However, the original RTA index is asymmetric and highly skewed. We therefore use a normalized index (RTA-1/RTA+1), which takes values between -1 and 1. Negative values indicate a competitive disadvantage of the firm in the technology, while positive values indicate an advantage. 10 The total population here consists of all firms having patents with Swedish inventors. 12

13 Core technologies are defined as those patent classes in the firms portfolio, which have Patent share and RTA above certain threshold values. The average patent share in our sample is used to draw the threshold value. Mirroring Granstrand et al. (1997), we put the threshold for the normalized RTA at We are here using the fourdigit level of the International Patent Classification (IPC4), which in our dataset adds up to 301 different patent classes. 12 We construct a dummy variable taking the value 1 if the focal patent is in the firm s core technologies and 0 otherwise. Thus, this variable is constructed to indicate whether a patent belongs to the firms core technologies. For our purposes, this variable can be interpreted as that when it takes the value 1 (i.e. belongs to the core), the focal patent is closely related to firms technology base and thus is the result of technologically local search (i.e. more exploitative than explorative): If it is an academic patent, it can then be seen as the result of Externally boundary spanning search, if it is not an academic patent it is the result of local search (cf. Rosenkopf and Nerkar, 2001). If the patent however does not belong to the core, then it is more distant from the technology base, i.e. it is the result of some degree of technological boundary spanning (related to technology diversification): If it is an academic patent it can then be seen as the result of radical search, involving both organizational and technological boundary spanning, otherwise it is the result of internal boundary spanning. We also construct a variable measuring the centrality of the focal patent s technological field, meaning the distance from the technology base. This variable is constructed as the share of firms patents in the technological class of the focal patent at the time of application (i.e. priority year), thus it is identical to the Patent share variable used to construct firms technological profiles. The idea behind this variable is that the higher share of the patent portfolio that is occupied by the focal patent s patent class, the closer the invention is to the firms technology base. 11 Although we follow these previous studies, there is some arbitrariness in regards to defining the threshold limits of the PS and RTA variables making up the technological profiles. We have therefore tested our models using different values, within reasonable ranges, of these two variables as the threshold, with overall consistent results. 12 We have conducted the same analysis using different levels of patent classifications (IPC3, IPC2, OST30), with consistent results. 13

14 To be able to provide some preliminary descriptive analysis, we operationalize the question of whether academic engagement affect large firms technology base over time in terms of the dynamics of firms patent portfolio. Put differently, we examine to what extent that the technologies in which firms involve academic inventors grow or diminish in importance in the patent portfolio over time, meaning whether or not the technologies patent share increases or decreases. Thus, for doing this analysis we change the unit of analysis from the patent to the patent class. We study the firms patent portfolio over 5-year periods (between 1985 and 2009) rather than tracking it from year to year, since patenting activities tend to fluctuate over time. To do so, we construct a variable that measures the growth (or decline) of patent classes share of firms patent portfolio in one period as compared to the previous period (using the Patent share variable). Moreover, we construct a dummy variable taking the value 1 if the firm has one or several academic patents in that technology class in the previous period. This variable is constructed in order to investigate the relation between the change of firms patent share of technology classes in one period and the involvement of academic inventors within those technology classes in the previous period. In Section 4, we present the results from a preliminary econometric analysis of the relation between academic inventors and the technology centrality of firms patents. As the dependent variable, we use the aforementioned Core dummy, indicating whether a patent belongs to the firms core technologies. Since the dependent variable is a dummy, we employ a probit model. In this preliminary model, we also include a few control variables. We include the total size of the firm s patent portfolio at the time of patent application (priority year). Finally, we control for differences across technological classes and priority years. 4. Results Here we report on the preliminary descriptive findings of our analysis. In Table 1, the descriptive statistics for the measures of technology centrality (Patent share) are presented. The table shows that firms academic patents on average have lower patent share, as compared to non-academic patents. 14

15 Table 1. Descriptive statistics of Technology centrality of firms patents 13 Non-academic patents Academic patents Mean Std. Dev. Min Max Mean Std.Dv. Min Max Technology centrality (Patent share) Observations Figure 2 presents the distribution of the patent share variable for firms academic and non-academic patents. As can be seen in the figure, academic patents are more concentrated in the lower span of the distribution, and less so in the higher span, than non-academic patents. This thus means that academic patents are more concentrated in technologies with low share of the firms patent portfolio. Overall, the descriptive results from Table 1 and Figure 2 indicates that academic inventors are involved in technologies that are more distant from the technology base, as compared to nonacademic patents Figure 2. Kernel density of Technology centrality (Patent share) for academic and non-academic patents 13 To test the significance of the difference in mean between the two groups of patents, we perfomed a Mann-Whitney rank sum test. This was done since the variable is highly skewed. The test showed that the difference in mean between the two groups is statistially significant (p<0.001). 15

16 The difference between academic and non-academic patents is in fact more pronounced when we examine the distribution of patents over firms core/non-core technologies. As can be seen in Figure 3, around 35 % of academic patents belong to firms core technologies, whereas non-academic patents are much more concentrated with around 50 % belonging to the core. Figure 2. Distribution of academic and non-academic patents over technological profiles (%) In Table 4, we present the results from a preliminary analysis of the relation between academic inventors and the technology centrality of firms patents, using the Core dummy as the dependent variable. Since the dependent variable is a dummy, we employ a probit model, and we include patent portfolio size, technological class and priority years as controls. As can be seen, in our model academic patents are associated with a lower probability of being in firms core technologies as compared to non-academic patents (p<0.001). 16

17 Table 4. Probit model for Core technology dummy Academic patent *** (0.0653) Patent portfolio size 0.001*** Technological classes Priority years (0.0007) Included Included Observations Pseudo R-squared LR Chi square Log-likelihood Standard errors in parentheses *** p<0.001 Taking these preliminary findings together indicates that when academics collaborate with firms on inventions they are, on average, involved in technologies that are more distant from the firms technology base as compared to their overall patent portfolio. Table 5. Descriptive statistics of Patent share increase No academic patents in previous period Academic patents in previous period Mean Std.Dev. Min Max Mean Std.Dev. Min Max Change in Patent share from previous 5 year period Observations Table 5 presents the descriptive statistics for the variable measuring the percentage increase in patent share for each patent class in firms patent portfolio over the 5-year periods. As shown, academic inventors are on average involved in patent classes that increase their share of firms patent portfolio in the following period (the increase in percentage is 36 %). This is substantially more so than what is the case for patent classes that do not involve academics in the previous period. 17

18 5. Conclusions This paper has presented a preliminary exploratory study of large firms engaging academic inventors in their invention process. A key issue in this context is whether and to what extent firms can bypass their tendency towards local search (cf. e.g. Rosenkopf and Nerkar, 2001) by involving academics to diversify their technology base. This paper provides a preliminary study of firm patents in order to explore this issue. By doing so, we can identify in which technological fields that the firms are already strong (Core profile), and hence where we would expect the tendency towards local search to be the most pronounced. We provided a preliminary descriptive analysis of firm-owned academic patents in order to examine the extent to which firms academic patents are found in firms core technology fields, and technology fields that become more central in the technology profile over time. We thus focus on two sub-questions: i) To what extent do large firms engage academics as inventors in their core technologies?; and ii) Does academic engagement affect (large) firms technology base over time, and if so how? For the first question, our preliminary findings show that firms academic patents are less concentrated within these large firms core technology fields as compared to their non-academic patents. Similarly, academic inventors are engaged in inventions, which are, on average, more distant from the firms technology base. These preliminary findings indicate that large firms may engage academic inventors to a large extent for diversifying their technology base. For the second question, the preliminary findings indicate that academic inventors are on average involved in technologies that in later periods become more important in the firms patent portfolio, in terms of increasing their share of the portfolio. This together with the previous finding suggests that large firms engage academic inventors in technologies in which the firms are seeking to move beyond local search and to enhance their efforts and competencies. While the existing literature provides two differing answers to the questions we address in this paper, our preliminary findings are in line with the claim of Santoro and Chakrabarti (2000): industrial firms do not typically use university relationships 18

19 to help strengthen and build core competencies, but rather that large firms engage with academics in order to build competencies in non-core technological areas. While their study was based on the perceptions of survey respondents, our study provides preliminary evidence that corroborates their findings by analyzing the actual outcome of university-industry collaboration. References Breschi, S., Malerba, F., Orsenigo, L., Technological Regimes and Schumpeterian Patterns of Innovation. The Economic Journal 110, Cohen, W., Nelson, R.R., Walsh, J., Links and Impacts: The Influence of Public Research on Industrial R&D. Management Science 48, Feller, I., Ailes, C., Roessner, J., Impacts of research universities on technological innovation in industry: evidence from engineering. Research Policy. Fleming, L., Recombinant uncertainty in technological search. Management Science 47, Fleming, L., Sorenson, O., Technology as a complex adaptive system: evidence from patent data. Research Policy 30, Fontana, R., Geuna, A., Matt, M., Factors affecting university industry R&D projects: The importance of searching, screening and signalling. Research Policy 35, Granstrand, O., Patel, P., Pavitt, K., Multi-technology corporations: Why they have "distributed" rather than distinctive core competences. California Management Review 39, Henderson, R., Clark, K., Architectural innovation: The reconfiguration of existing product technologies and the failure of established firms. Administrative science quarterly 35, Henderson, R., Jaffe, A., Trajtenberg, M., Universities as a Source of Commercial Technology: A Detailed Analysis of University Patenting, The Review of Economics and Statistics 80, Lissoni, F., Sanditov, B., Tarasconi, G., The Keins Database on Academic Inventors: Methodology and Contents. CESPRI Working Paper 181. Ljungberg, D., McKelvey, M., What characterizes firms' academic patents? Industry & Innovation, Forthcoming. Mansfield, E., Academic research and industrial innovation. Research Policy 20, Mansfield, E., Academic research and industrial innovation: An update of empirical findings. Research Policy 26, March, J., Exploration and exploitation in organizational learning. Organization Science. Mowery, D., Sampat, B., Ziedonis, A.A., Learning to Patent: Institutional Experience, Learning, and the Characteristics of US University Patents After the Bayh-Dole Act, Management Science 48, Nelson, R.R., Winter, S., An evolutionary theory of economic change. Harvard University Press, Cambridge. 19

20 Nickerson, J., Zenger, T., A knowledge-based theory of the firm: the problemsolving perspective. Organization Science 15, Nooteboom, B., A cognitive theory of the firm learning, governance and dynamic capabilities. Edward Elgar Publishing. Perkmann, M., Walsh, K., The two faces of collaboration: impacts of university-industry relations on public research. Industrial and Corporate Change 18, Rosenkopf, L., Almeida, P., Overcoming local search through alliances and mobility. Management Science 49, Rosenkopf, L., Nerkar, A., Beyond local search: Boundary-spanning, exploration, and impact in the optical disk industry. Strategic Management Journal 22, Santoro, M., Chakrabarti, A., Firm size and technology centrality in industry university interactions. Research Policy 31, Scherer, F., Firm Size, Market Structure, Opportunity, and the Output of Patented Inventions. The American Economic Review 55,