Mapping the field: a bibliometric analysis of the literature on university–industry collaborations

3.1 Co-citation analysis results

The results of co-citation analysis are illustrated in Fig. 1. The visualised bibliometric network represents four interconnected clusters with unique labels that are based on the cited references from the initial dataset of 435 publications. The original dataset consisting of 435 articles published in the period of 2011–2016 refers to 14,705 references that are standardised to the same format for further analysis. However, in the co-citation analysis, we include publications with at least ten citations to ensure an analysis of an essential literature, avoiding the potential risk of overly complicating the interpretation (Kovács et al. 2015). This restriction leads to 245 references meeting the threshold. In accordance with the goal to examine the evolution of the U–I research field and the corpus of references used for co-citation analysis, we also include the cited literature published before 2011.

Open image in new window

Each scientific publication depicted in the visualised bibliometric network represents a unique cited reference that is clustered based on the likelihood to be cited in combination with other items. Items that are assigned to the same cluster share a higher probability to be cited together by another publication. Colours indicate the clusters and articles belonging to them. Moreover, a specific weight is assigned to each item based on a total link strength of the publication and the number of citations it has received. For a better overview, publications with higher total link strength are visualised in a larger size (van Eck and Waltman 2010).

Based on the results of the co-citation analysis, we retrieved the information necessary to conduct an additional examination of the journals interested in publishing U–I research and the keywords associated with the analysed publications. The results show that the three leading peer-reviewed journals (Fig. 2) that published the highest number of contributions to the U–I literature (period of 2011–2016) are Research Policy (95 publications), followed by the Journal of Technology Transfer (16 publications) and Management Science (9 publications). Moreover, we measured the term frequency of keywords assigned to the journals to examine the most important keywords in the cited references of the initial U–I literature dataset (Fig. 3).

Open image in new window

Open image in new window

Based on the visualisation of similarities (VOS) approach and the co-citation analysis technique, the current body of U–I research was divided into four thematic clusters and labelled based on the authors’ independent content analysis. In addition, it needs to be acknowledged that—while the authors’ attempted to define clusters in a manner that would reflect the entire scope of the U–I literature—the foci of the articles included in the analyses are of a broad variety; therefore, certain elements cannot be fully elaborated on.

3.2 Bibliographic coupling results

The results of bibliographic coupling that outline the current research perspectives in the U–I research field are illustrated in Fig. 4. Similar methodological steps are applied to the interpretation of bibliographic coupling results. Nevertheless, the implemented differences should be clearly outlined. We apply no restrictions concerning the number of citations to grasp the entire dataset with research articles published from 2011 to 2016. As noted previously, this choice has been made to complement the previous research and develop new insights based on previous reviews. From a total sample of 435 articles included in the analysis, 402 articles were clustered and included in the illustrated bibliographic network. Excluded articles represent a sample of publications with no shared references with other articles in the dataset and therefore cannot be connected. In addition, two clusters were composed of 3 and 2 articles, respectively; due to the low applicability, these were excluded from further analysis. Thus, the final sample of analysis consists of 397 clustered articles.

Open image in new window

Based also on the bibliographic coupling results, we retrieved the corresponding information enabling an analysis of the journals publishing the relevant literature and the keywords that are assigned to these publications. Figure 5 represents the journals publishing most research on U–I collaborations, based on the number of scientific publications per journal in the period of 2011–2016. The three leading journals are Research Policy (42 publications), followed by Journal of Technology Transfer (39 publications) and Scientometrics (32 publications). A term frequency analysis was performed revealing the most occurring keywords assigned to the publications (Fig. 6).

Open image in new window

Open image in new window

Applying the visualisation of similarities (VOS) approach and the bibliographic coupling technique, the existing research was divided into six thematic clusters and labelled based on the authors’ independent content analysis.

4.1 The individual in university–industry collaborations

On the individual level, with regard to the social relations perspective, U–I scholars recognise the importance of the relationships between the individuals involved in the collaborative U–I activities. While the recent research endeavours must be applauded, there is still a long way to go on the question of what individual-level characteristics and motives should be possessed by academics and industrial partners to successfully collaborate, and particularly to manage the relationships over time to achieve strong outcomes (Filippetti and Savona 2017; Maietta 2015). Given that the characteristics and motives of academics and industrial partners differ when it comes to U–I collaborations, in addition to within-group variations (Link 2015), it is likely that the mechanisms for governing U–I relationships are highly varied in nature. In particular, future studies are recommended to investigate which types of governance mechanisms are effective depending on partners’ motives and characteristics. For example, firms investing in university research projects may be interested in an alignment of goals as well as commitment between the collaboration partners (Mora-Valentin et al. 2004). A successful alliance needs goal-oriented management (Kale and Singh 2009), so that from the firm perspective it may be beneficial to apply control mechanisms. However, as researchers tend to have a high need for autonomy in setting research directions and managing their research in U–I projects (Zalewska-Kurek et al. 2016), a balance between control-based and more open or heterarchical governance may need to be negotiated to fit the expectations and needs of both partners. With regard to the partners’ characteristics, for firm partners managing a portfolio of relationships, weak ties/loosely coupled relations may be the preferred mechanism to create spaces of dynamism and thereby generating profits from the engagement in a variety of U–I collaborations. Weak ties are characterised by low interaction frequency and can enable access to new, non-redundant knowledge (Granovetter 1973). However, stronger ties might want to be established on behalf of the firm partner when interacting with a specific star scientist or a group of scientists who significantly contribute to the firm’s knowledge base. In turn, academics—due to their dual, complex nature of employment as researchers at universities and as self-employed within U–I collaborations—may want to build strong ties characterised by high interaction frequency, and social proximity to create spaces of stability. In addition, stronger ties between the U–I collaboration partners might be more effective for generating substantial knowledge-related benefits, while weaker ties could be more suitable in service-provision partnerships for generating financial benefits on the part of research organisations (Arza and Carattoli 2017). It would be intriguing to analyse which type of social capital comes to the fore in U–I collaborations, why, and with what effects. Moreover, the governance mechanisms may vary according to the stage of the cooperation. While there initially may be a strong focus on formal governance mechanisms to reduce uncertainty within the collaboration, some partners may put more emphasis on informal mechanisms such as trust as the U–I collaboration gradually develops—which might explain why some U–I collaborations are more or less successful than others. This is because in joint R&D, neither the goals nor the contributions of the partners can be defined completely ex ante and thus hierarchical tools of control and sanction (e.g., Adler 2001; Williamson 1991) fail to coordinate innovative cooperative behavior over time (Hatak et al. 2015). Here, we encourage researchers to extend current research by engaging in cross-stage studies. Additionally, we must acknowledge that the U–I relationships are shaped by institutional environments in which the academics and the industry partners are embedded, thus affecting the (expected) outcomes from joint R&D activities (Frasquet et al. 2012; Hemmert et al. 2014; Morandi 2013). Thus, it is necessary to consider the institutional context as a moderator in the relationship between governance mechanisms and outcomes of U–I collaborations. Moreover, future studies could focus on optimising the management of a portfolio of incentives and initiatives that foster U–I linkages during different evolutionary stages of partnerships (Ankrah et al. 2013; Plewa et al. 2013b). Correspondingly, there is a need to investigate which incentives not only enhance both researchers and industry partners’ willingness to engage in U–I partnerships, but also foster the development of long-lasting strategic alliances. Understanding the mechanisms enabling to maintain long-term U–I relations would improve the viability of the knowledge transfer, forming the basis for further R&D projects with reduced barriers. According to Bienkowska and Klofsten (2012), fostering the development of relationships among the partners could be achieved by promoting commercialisation of research results (CRR) skills and providing new opportunities for PhD candidates through courses, graduate schools and interactive research designs. At the same time the question remains of how U–I collaboration can be made attractive for partners from the industry by incentives other than funding schemes.

Research focusing on Academic Entrepreneurship explores the characteristics and motives of the academic entrepreneur, the entrepreneurial university, and the context in which the actors of academic entrepreneurship are embedded, and their effects on spin-off success. The central element connecting these layers of academic entrepreneurship is the entrepreneur (Perkmann et al. 2013; Tartari and Breschi 2012). The increasing resources devoted to understanding the behavioural outcomes associated with entrepreneurs’ identity in general calls for U–I research that identifies different types of identities among academic entrepreneurs in particular. This is an interesting avenue as academics need to integrate their academic self-concept and their entrepreneurial self-concept, resulting in either hybrid identities or pure identities, which, in turn, offer explanations of the commercial behaviour of academics (Abreu and Grinevich 2017). In this regard, it is important to explore the intrinsic and extrinsic motivations driving entrepreneurial behaviour of academics, such as time spent and duration of engagement capturing behavioural intensity and persistence (Lam 2011). To analyse behavioural intensity and persistence, one might use experience sampling methodology (Uy et al. 2010). As shown by research on the motivations of academics to start a spin-off, their motives differ (Fini et al. 2009; Hayter et al. 2016; Rizzo 2015) so that their identities can also be expected to be heterogeneous (Clarysse et al. 2011). Another interesting research avenue that we propose would be to examine the university spin-offs’ (USOs) failures and the associated failure factors on the level of the individual actors. While academic entrepreneurship has a long tradition of examining success factors, it might actually be the focus on failure factors that has the potential to provide counterintuitive and therefore relevant new insights. Furthermore, a relatively new stream of literature points to the impact of students, recent graduates and young academics on regional innovative development. Identifying the factors and conditions that stimulate these actors to increasingly engage in the commercialisation activities of their research and understanding the scope of this ecosystem would be of significant value (Laursen et al. 2011; Astebro et al. 2013; Wright et al. 2017). Such studies might provide answers to questions that have been only partially addressed: What are the characteristics of the graduates that start companies (e.g., Larsson et al. 2017)? Under what conditions do they start their companies? Furthermore, what types of entrepreneurship education (e.g., regular courses, bootcamps, start-up weekends, teaching methods directed towards creating an entrepreneurial mind-set, attitude and orientation) actually increase students’ and recent graduates’ entrepreneurial preferences and their subsequent translation into start-up actions? For example, is the lean start-up methodology more effective the traditional business methodology? Experimental studies might enable researchers to disentangle the various influences of entrepreneurship education at the different sub-stages of the entrepreneurial process, such as the development of intentions, start-up actions and young venture development. In addition, exploring the effects of start-up activity by recent university graduates, PhDs and post-docs on the human capital movement in the region and the R&D performance of the region, respectively, are crucial research dimensions (Laursen et al. 2011). Recently, an excellent outlook on future academic entrepreneurship research has been drawn by Siegel and Wright (2015), to which we refer the interested reader. The authors discuss in more detail some of the abovementioned avenues, among others.

4.2 The organisation in university–industry collaborations

On the organisational level, with regard to the academic entrepreneurship perspective, the U–I literature can be consolidated into the analysis of the characteristics of universities engaging in joint research activities, characteristics of firms seeking to collaborate with academic partners and the characteristics of the collaboration process itself. There are studies indicating why firms and universities engage in joint projects, but research on the determinants of private and public gains of university involvement in research partnerships with industrial partners is still at a nascent stage (Link 2015). In this regard, much research attention needs to be devoted to the heterogeneity of U–I partners. It is likely that the antecedents and consequences of academic entrepreneurship differ when taking account of, for example, the firm size, age, and competitive strategy of the industrial partner, and the size, age/ranking, type of university. On a related note, research on the impact of a specific department’s excellence regardless of the university ranking on decisions to engage in partnerships on regional, national or international levels could yield interesting findings (Laursen et al. 2011). Furthermore, while there have been calls for research on firms’ strategies when searching for and selecting academics as collaboration partners (see e.g., Perkmann and Walsh (2007), the industry’s criteria driving the choice of a specific academic partner for a collaborative research project remains an underexplored topic (Link 2015). Special attention needs to be devoted to the strategic and cultural fit between university and industry partners for understanding the successful organisation and management of their collaborative projects. Finally, new success factors of university entrepreneurship projects should be examined in the future studies, which could be achieved by the systematic evaluation of failed projects and collaborations, and particularly the factors of failure at different stages of project life-cycle (Bozeman et al. 2013; Guzzini and Iacobucci 2017).

While the distance perspective offers an increasing understanding of the outcomes of partner spatial, cognitive, organisational, social and institutional proximities, the multi-dimensionality and broad scope of this perspective suggest further research opportunities (Crescenzi et al. 2017; D’Este et al. 2013). The academic understanding of U–I partnership’s evolvement and outcomes may benefit from exploring the interplay of different types of proximities and associated dynamics (Muscio and Pozzali 2013; Presutti et al. 2017). Here, future studies could further examine whether and how a specific type of proximity can foster the development of other proximity types—for example, spatial proximity may lead to social proximity—as well as whether one type of proximity may replace another depending on the context in which the U–I collaboration is embedded (Crescenzi et al. 2017). However, as the capturing of direct and particularly indirect effects of partner proximity remains challenging, researchers are encouraged to develop alternative outcome indicators to reliably estimate the effects of partner proximity in the context of joint R&D activities (D’Este et al. 2013; Fukugawa 2013). Moreover, our review points to the benefits of focusing the distance perspective research on partner centrality and closeness. For instance, what is the role of central partners within the network regarding the joint project success (von Raesfeld et al. 2012)? Specifically, future research could examine how central partner presence in collaborative projects affects the capability to generate innovative outcomes, potentially through the optimisation of accessible knowledge and technology inputs (Qi Dong et al. 2017). In this regard, the role of other factors complementing or mediating the impact of partner proximity is an interesting area for future studies. For example, the availability of research resources, the partner technological complementarity and the level of absorptive capacity in conjunction with partner proximity remain important topics for future studies (Bishop et al. 2011; Laursen et al. 2011; Petruzzelli 2011). Another unexplored aspect of the proximity stream within the U–I literature is the extent of industry involvement in the creation of university spin-off (USOs). Academic entrepreneurship has mainly been supported by the university and government (Meyer 2003; Fini et al. 2011; Grimaldi et al. 2011); however, it can be also supported by the (local) industry. Simultaneously, firms may even decide for locations that are geographically close to research organisations to benefit from knowledge spillovers (Audretsch et al. 2005). Anecdotal evidence suggests that industry partners geographically close to the university provide academic commercial ventures with access to facilities and knowledge from their business development units (private business clusters or incubators) rather than with substantial capital funds. While there is research on the shaping of industries’ support in the early stages of spin-off development (e.g., Zhang 2009), the mutual resource flows between USOs and their industry partners over time and in conjunction with partner proximity have received little conceptual and empirical attention.

4.3 The institutions in university–industry collaborations

On the institutional level, the literature points to the economic and societal impact generated by U–I collaborations on regional and national levels as well as to an increasing importance of policy developments that would maximise the effectiveness of U–I interactions (Filippetti and Savona 2017). In the last decade, different configurations of Triple Helix interactions that frequently form a complex system of inter-dependent relations have been developed (Lei et al. 2012). Based on our review, future research embedded in the ecosystem perspective would benefit from examining what types of Triple Helix network configurations are most effective, considering institutional differences across countries and regions (Etzkowitz 2011; Guan and Zhao 2013; van der Valk et al. 2011). In this regard, prior research calls for the development of performance indicators that would reliably predict what configurations are necessary to optimise collaborative project success between university–industry–government (Feller 2017; Leydesdorff and Rafols 2011). Pursuing this research avenue will become increasingly relevant due to the development of an e-government domain (Khan and Park 2011). In addition, given the recent increase in opportunities (e.g., university/public funding support for U–I projects) and the threats (e.g., weak IPR protection) in the context of (international) U–I collaboration, future research is recommended to extend the framework of university–industry–government relations towards Quadruple Helix, thus expanding the facets of U–I collaborations that are examined (Leydesdorff 2012; Marcovich and Shinn 2011). Moreover, a relevant exploration of network configurations needs to account for differences between industry sectors and technology fields (Guan and Zhao 2013; Kwon et al. 2012) as well as between scientific fields. It can be expected that all scientific fields have an impact on economy and society; however, social sciences have largely been left out in U–I collaborations research despite their contributions to the competitiveness and innovativeness of the economy by providing e.g., new research tools and interventions in conjunction with consultancy, knowledge on the effectiveness of management approaches, highly skilled and entrepreneurially educated graduates among other outcomes. More research could be done on the question of how and with what effects knowledge is transferred from social sciences to the society at large.

With regard to the policy implications perspective, research argues that university–industry–government interactions can be shaped by civic involvement and their characteristics, in addition to cultural and social capital endowments (Lombardi et al. 2012). Thus, it is worthwhile to investigate how changes in the local institutional environment foster the engagement of academics in collaborative research projects and university spin-off creation (Fini et al. 2011). Conversely, it would be interesting to examine how U–I collaborations affect policy development, requiring approaches that are multi-disciplinary and historical in nature. In relation to the incentives that could stimulate engagement in U–I collaborative projects on the part of academics, research should consider that collaboration with industry does not necessarily foster success in academia. As indicated by Lee and Miozzo (2015), the engagement in industry-involved projects by physicists and engineering Ph.D. researchers is negatively related to their career in academia. It is therefore not only up to the university but the entire science system including policy makers to come up with an incentive scheme specifying how and when academics can be promoted if they engage in the third mission of the university. This issue need to be further explored by research, outlining the characteristics and conditions of best practices.

The process perspective (interaction process and knowledge transfer cluster) of U–I collaborations is not a core research stream; however, there is a strong need for future examinations, especially if we want to understand the complex processes of interaction between academia and industry, resulting in a long-term exchange of knowledge and technological resources at all levels of U–I collaboration (Gilsing et al. 2011). In line with McMullen and Dimov (2013), there are a number of conceptual questions that require answers if a process perspective to U–I collaborations shall gain momentum: What is the U–I collaboration journey, when has this journey started, when has this journey concluded, does it require particular interactions to progress; and what remains constant and what changes throughout the process of interaction between U–I partners? Longitudinal studies could help to answer those questions and capture the depth and breadth of interaction cycles, beginning with an initiating action and both the academics’ and industry partners’ reactions. Finally, our review points to the benefits of extending the level of analysis of future U–I interaction process studies beyond the individual level to ask—among other questions—whether industry partners in the same industry react in similar ways, how interaction is developed at higher levels, that is between the firm, the university and the government, and how such interactions shape the dynamics of academia, the industry and the institutional environment.

4.4 Research limitations

Inevitably, this study also has certain limitations paving the way for several extensions of the bibliometric analysis of U–I collaboration research and additional analyses. First, this study investigates a specific dataset on U–I collaborations. However, the results show that each of the identified perspectives in this paper could be investigated independently. Thus, it is necessary to extend the scope of the U–I literature among the identified research perspectives. Additionally, using other keywords for data collection may lead to broader research directions. Second, the main goal of this study was to grasp the latest research advancements. Future studies could explore the interplay of our findings in comparison to the entire field while imposing no time limitations. Third, our dataset is focused on the most recent publications; therefore, the weight of the publications used for creating the clusters of literature will change according to the citation patterns of this research field. Finally, imposing a threshold on the number of citations for co-citation hinders an analysis of the entire literature, which also ignores potentially relevant but primarily recent publications. An investigation of different datasets compiled on the basis of other keyword combinations and extended period of publication time could enable us to identify additional elements of U–I collaborations and academic entrepreneurship and establish new interlinkages across the U–I ecosystem.

4.5 Conclusion

This study examined the nature of the U–I collaborations research using both quantitative and qualitative methods of analyses, concluding that the research field is complex and multi-faceted. Based on co-citation analysis, we analysed the evolution of the research field and identified fundamental research themes. By using the method of bibliographic coupling, we were able to investigate current and emerging topics of interest in the study of U–I collaborations. Based on our comprehensive content analysis, we were able to identify linkages between the different perspectives adopted in U–I collaborations research, thus contributing to integration of diverse research streams and a clarification of the scope of the U–I collaborations literature. The findings of our bibliometric study suggest that the U–I collaborations research can be seen as a multi-layered ecosystem consisting of interconnected perspectives on individual, organisational and institutional levels. Our study concludes by offering a nuanced research agenda with valuable implications for academics and policy makers.