STRONG TIES BETWEEN ACADEMIC AND CORPORATE RESEARCHERS: FOCUS ON TRIGGERS FOR TIE FORMATION

Introduction

Innovation occurs from new combinations of existing knowledge, and external knowledge search and acquisition are the first critical steps towards new combinations of knowledge (Cohen and Levinthal, 1990; Yu et al., 2022). While having networks across organisations is widely recognised as important for innovative firms, most studies have focused on firm- or regional-level networks, and only a very limited number of studies have targeted interpersonal relationships across organisational/regional boundaries (Huber, 2012a; Huber and Fitjar, 2016). Among such scarce studies, some scholars have analysed the geography of personal networks and found that they are mostly local but extend to the national and global levels by overcoming the geographic distance between actors (Fitjar and Huber, 2015; Murakami, 2019; Xu et al., 2019). In addition, Xu et al. (2019) argued that triggers for tie formation, knowledge function, and geographic scope differ by type of personal network.

On the other hand, there is very little research on personal networks that transcend the institutional distance between academia and industry. As discussed in the second section, there is institutional distance between them, which makes it difficult for knowledge exchange networks to be formed. However, there are great expectations that heterogeneous knowledge linkages through academia–industry interactions will result in innovative performance. In fact, Crescenzi et al. (2017) argued that University–Industry collaborations are less likely to occur, but once they are established, they tend to create patents of greater impact and more general technological applications.

The strength of ties is an analytical perspective in network studies. While strong ties contribute to innovation, as discussed in the second section, Murakami (2023) revealed that not only is the knowledge exchange network among academic and corporate researchers small but also the ties are likely to be weak compared with ties connecting academic researchers. Nevertheless, we can expect that there are both strong and weak ties even among University–Industry personal networks. It would be helpful to learn how strong ties can be formed among academic and corporate researchers to promote the absorption of external knowledge, but no such study has been conducted. This study is an attempt to fill this gap.

According to Dahl and Pedersen (2004), who examined the origins of 346 engineers’ informal contacts in the wireless communications cluster in Denmark, 66% of the contacts were former colleagues, 50% were classmates, 47% were private friends, and 8% were others. Personal networks across organisational boundaries are more likely to form within a cluster/region (Dahl and Pedersen, 2004; Singh, 2005; Head et al., 2019), and within a cluster, there are many opportunities to meet through trade fairs, meetings, social activities, etc. (Saxenian, 1994). The context in which people interact and share tacit knowledge through informal communication is called “Ba”, and its sharing is useful for network formation (Nie et al., 2010). For example, Huber (2012a) analysed the personal networks of 105 R&D workers in hardware and software industries in Cambridge and reported that 82.1% of the relationships between them and their most important knowledge contact began in face-to-face situations. All former colleagues, classmates, and acquaintances from conferences, seminars, and social activities are relationships that shared a “Ba” in the past. The differences in “Ba” at the time of network formation may affect the strength of ties. This is because Granovetter (1973, p. 1361) defined tie strength as “a (probably linear) combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie”, and there seems to be difference in the level of the amount of time and the intimacy between ties formed at a conference and those formed as colleagues or schoolmates, for example.

In addition to the perspective of shared “Ba”, there is also the perspective of the presence of intermediaries in the analysis of the strength of ties (Staber, 2011): the introduction by a third researcher or an industry–academia intermediary organisation. The contribution of the latter is particularly worthy of consideration in the context of the difficult formation of networks between industry and academia. The knowledge and technology created at universities are expected to be jointly developed by universities and firms to supply new products, services, and systems or to improve the efficiency of production processes, which possibly enhances economic growth and international competitiveness. Thus, various countries have promoted industry–academia collaboration (Rham et al., 2000; Crescenzi et al., 2017; Gallagher et al., 2023), and intermediary organisations have been established in universities (Siegel et al., 2007; Boardman and Gray, 2010; Lee, 2014; Villani et al., 2017). Whether the ties formed through these intermediary organisations are stronger than those formed through personal relationships between researchers is yet unclear, but it is important to address this issue from a policy evaluation point of view. Therefore, this study analyses the strength of network ties between academic and corporate researchers from the perspective of triggers for network formation, aiming at drawing insights into management and public policies for the formation of strong ties.

This study considers five types of triggers for network formation based on prior studies: “university acquaintance”, “colleagues and mentorship at work”, “introduction by peer researchers”, “acquaintances from academic conferences (including seminars)”, and “introduction by mediating organisations”. The mechanism that relates these triggers and the strength of ties is trust building and development, which is the driver of knowledge exchange between industry and academia. As described below, there are two types of trust: cognition-based and affect-based. Depending on the triggers, there might be differences in the length of time the parties spend together and the presence or absence of the third party’s intervention, which might influence these two types of trust.

The data for my empirical analyses are 508 ties of personal knowledge exchange networks connecting academic and corporate researchers in Japan, which were collected via a survey. Following Fitjar and Huber (2015), personal networks are a set of individuals and their personal relationships that are informal but that might be embedded in coexisting formal relationships. Japan has the third largest population of researchers in the world (National Institute of Science and Technology Policy (NISTEP, 2023), and industry–academia collaboration has been promoted by the government. Focusing on a single country is helpful for controlling the cultural and institutional differences among countries, although there may be limitations in generalising research results.

This paper is organised as follows. The following section first summarises previous studies on the strength of ties and the barriers to and drivers of academia-industry network formation and then presents the hypotheses of this study. In the Method section, the data and empirical model are described. After presenting the results of the empirical analysis and examining the hypotheses in the Result section, the Discussion section interprets the results, discusses the relationships between the triggers and the strength of ties, and suggests management and public policies for the formation of strong ties between academic and corporate researchers. The last section concludes this study and refers to future research.

Conceptual Framework

Strength of ties

This study is based on network theory. In empirical studies, tie strength is measured by the strength of the dyadic relationships or network structure. The former measures tie strength by the frequency of interactions/communication and degree of intimacy between two actors (Hansen, 1999; Reagans and McEvily, 2003; Levin and Cross, 2004; Rost, 2011) or the number of past copatents (Ponds et al., 2010; Lin and Wang, 2019). From the latter point of view, tie strength is measured by the number of intermediaries between knowledge senders and recipients, which is called the geodesic distance of the shortest path. If two actors are directly connected, the distance is the shortest; the greater the number of intermediaries between two actors is, the longer the distance is, and a tie with a long distance is a weak tie (Hansen, 1999; Singh, 2005; Cassi and Plunket, 2015; Crescenzi et al., 2017).

An interesting question is whether strong or weak ties are more useful for acquiring knowledge. When tie strength is measured by the strength of the dyadic relationship, the two actors connected by a strong tie have strong emotional attachment, mutual trust, and motivation to support each other (Powell and Grodal, 2005; Reagans and McEvily, 2003; Hansen, 1999; Levin and Cross, 2004), which encourages the sharing of complex and proprietary knowledge (Rost, 2011; Cassi and Plunket, 2015). Especially in uncertain situations, strong ties based on mutual trust are expected to promote knowledge exchange (Krackhardt, 1992). Furthermore, actors connected by strong ties are likely to have common ways of thinking and communicating, which also makes knowledge exchange easier (Levin and Cross, 2004).

Paying attention to dyadic social distances within a network architecture, on the other hand, weak ties are likely to be sources of novel knowledge because they can connect socially distant actors (Granovetter, 1973; Powell and Grodal, 2005; Levin and Cross, 2004). In contrast, strong ties are not expected to be a source of new knowledge because they are likely to be embedded in a cohesive network that comprises redundant ties (Powell and Grodal, 2005).

However, strong ties embedded in a network with redundant ties are more useful than weak ties when the transferred knowledge is complex/tacit or when one needs confirmation and reinforcement from multiple sources before taking action (Hansen, 1999; Centola and Macy, 2007). Rost (2011) showed through analyses of key inventors of the German automotive industry that weak network architectures with structural holes proposed by Burt (1992) cannot create innovation without strong ties, as measured by the frequency of knowledge exchange about professional issues, whereas strong ties have a high likelihood of invention even when embedded in a redundant network. The effects of strong ties in less cohesive networks have also been examined. For example, McFadyen et al. (2009) analysed networks of university research scientists in biomedicine and concluded that researchers who maintain mostly strong ties with research collaborators who themselves comprise a sparse network have the highest probability of creating impactful knowledge.

With respect to the effects of strong ties on innovation, Fritsh and Kauffeld-Monz (2010) argue that although Granovetter posits that new information is obtained mainly through weak ties rather than through strong ties, adopting this argument in the context of innovation activity may be problematic. They state that whether strong or weak ties are more desirable depends on the nature of the subject that must be transferred. Strong ties are better suited to an exchange of complex knowledge, whereas weak ties could be more beneficial for information search. From the above studies, we conjecture that industry–academia networks are scarce and have a less cohesive architecture, but strong ties between academic and corporate researchers can contribute to knowledge transfer and subsequent innovative knowledge creation.

Barriers and drivers of university–industry personal networks

Differences between industry and academia are often noted (Bruneel et al., 2010; Sauermann and Stephan, 2010; Albats et al., 2020; O’Dwyer et al., 2023), which might be a barrier to the formation of networks between academic and corporate researchers. The industry mainly conducts applied and development research aimed at the commercialisation of research outcomes. Academia, on the other hand, focuses more on basic research, the results of which are public goods and have no direct commercial value. This difference corresponds to the distinction between Mode 1 and Mode 2 knowledge production presented by Gibbons et al. (1994). Mode 1 knowledge is generated in the cognitive context of a particular discipline, and individual creativity is emphasised for development. Problems are set and solved in a context governed by the interests of an academic community. Mode 2 knowledge, on the other hand, is created in transdisciplinary social and economic contexts by a heterogeneous set of practitioners with practical goals and carried out in the context of application. Mode 2 knowledge is characterised by transdisciplinarity and heterogeneity.

Mode 1 knowledge has characteristics of public goods and should be widely utilised. Academic researchers cannot profit from the exclusive possession of created knowledge, and the recognition of their peers in the scientific community is their main incentive for research. Thus, they disclose their research results through publications and conference presentations. On the other hand, corporate researchers, who are conducting applied and development research with the potential for commercialisation, are more cautious about publication. They are motivated to possess the economic value of new knowledge to gain competitive advantages in the market. Furthermore, in academia, where individual creativity is emphasised in the production of Mode 1 knowledge, individual researchers have a high level of discretion over choosing research themes and partners, how to approach the themes, disclosure of research results, etc., whereas private firms, where teams of experts from various disciplines are required to achieve practical goals under time constraints, are governed by a hierarchical organisation.

Albats et al. (2020) summarise these differences in industry and academia into four barriers: “connection” (difficulties in finding and connecting to the right partner), “resources” (lack of funding, human resources, knowledge, etc.), “culture” (motivation, modes of communication, language, time horizon, and bureaucracy), and “internal organisational differences” (disagreements on intellectual property rights and limitations to a firm’s capacity to acquire knowledge from universities). However, there are also drivers to overcome them and proceed with collaboration. According to Albats et al. (2020, p. 2), the core drivers are “the availability of complementary resources (funding, human resources, knowledge, etc.) and relational drivers (trust, commitment, shared goals, and balancing of differing expectations)”. In other words, the existence of complementary resources that are desired even at the cost of efforts to overcome the barriers and the establishment of relationships that reduce the barriers are the driving forces behind their collaboration.

However, complementary knowledge that researchers from industry and academia would like to exchange may not exist because they have different interests, as described above. In other words, the creation and maintenance of knowledge exchange networks and strong ties connecting academic and corporate researchers depend in part on the existence of knowledge to be exchanged. On the other hand, relational drivers concern the tie strength defined by Granovetter (1973); mutual trust between academic and corporate researchers creates a strong tie and becomes a relational driver, which promotes knowledge exchange between industry and academia, even if the knowledge exchanged is not necessarily complementary. Bruneel et al. (2010) found, from a survey of firms that have been actively engaged in collaboration with universities, that interorganisational trust is one of the strongest mechanisms for lowering barriers. Similarly, Galán-Muros and Plewa (2016) showed, from a survey targeting European academics, that personal relationships (trust, commitment, shared goals, etc.) are the most important factor in facilitating university-business cooperation.

Basically, knowledge exchange through personal networks is a social exchange and is thus based on trust. Social exchange is the exchange of services that are internal rewards to others, such as pleasure and gratification (Blau, 1964). Whereas economic transactions are either immediately implemented or based on formal specific contracts, social exchange is based on trust in others’ fulfilment of obligations and reciprocation. There are two forms of interpersonal trust in general: cognition-based and affect-based trust (McAllister, 1995; Levin and Cross, 2004; Gallié and Guichard, 2005). The former involves trust in the competence and responsibility of partners, and the latter involves emotional bonds (reciprocated interpersonal care and concern for the welfare of partners). The latter is derived from frequent interaction/communication that provides the opportunity to gauge partners’ beliefs and attitudes and may vary with the closeness of network ties (Reagans and McEvily, 2003; Ren et al., 2016). In contrast, the former can be built on the basis of evidence such as reliable performance records and professional credentials, which are good clues for evaluating the competence of partners. Gallié and Guichard (2005) noted that the former can be developed through reputation, papers and conference presentations. In other words, cognition-based trust does not necessarily result from the length of a relationship. McAllister (1995) stated that cognition-based trust is considered “more superficial and less special” than affect-based trust and that “for working relationships among managers, some level of cognition-based trust may be necessary for affect-based trust to develop” (p. 30). He further stated that once a high level of affect-based trust develops, the foundation of cognition-based trust may no longer be necessary.

Relationship between triggers for tie formation and the strength of ties

This study analyses the strength of personal ties between academic and corporate researchers from the perspective of triggers for tie formation. As shown in Fig. 1, I assume that the main mechanism linking triggers for tie formation and the strength of ties is “trust”, which is a driver of knowledge exchange between industry and academia. The triggers for tie formation that this study focuses on are “university acquaintance”, “colleagues and mentorship at work”, and “acquaintances from academic conferences (including seminars)”, which prior studies have referred to, as well as “introduction by peer researchers” and “introduction by intermediary organisations”, which a third party intervenes in. Depending on these triggers for tie formation, the length of time that the parties spend together as well as the frequency of interaction and communication in a “Ba” may differ, which may influence the development of affect-based trust. The presence or absence of third-party referrals also differs by triggers for tie formation. The introduction by a third party ensures the partner’s ability and commitment, which is expected to result in cognition-based trust.

In this study, “acquaintances from academic conferences” of the five triggers is set as the base category because the time the actors spend together is short and there are no third-party intermediates. The length of time shared with teachers and friends of school age is longer than that of encounters at academic conferences. We can infer that through repeated interactions in various collaborative/group settings at school age, one can gain accurate perceptions of one’s friend ability and aptitude and that affection develops between them. Therefore, I propose Hypothesis 1.

Hypothesis 1: A tie with a university acquaintance is stronger than a tie with an acquaintance from an academic conference.

Similar reasoning holds for previous colleagues. The length of time they work together may vary, but at the very least, it is likely to be longer than that of encounters at academic conferences. Interactions in working together towards the same goal may help build cognitive and affective trust. Therefore, Hypothesis 2 is presented.

Hypothesis 2: A tie with an acquaintance from past work is stronger than a tie with an acquaintance from an academic conference.

With respect to “introduction by peer researchers”, we can expect that the person who introduces the partner ensures the partner’s ability and personality and that cognitive trust is likely to be built. A third-party introduction also forms a triad closure. When X and Y exchange business cards in person at an academic conference, the relationship may ultimately be a two-party relationship. On the other hand, when X and Y become acquainted through an introduction by Z, a triad is formed, and the relationship between X and Y is not easily broken. Granovetter (1973) stated that the most unlikely triad is that in which A and B are strongly linked and A has a strong tie to C, but there is no tie between C and B. Therefore, if Z has strong ties with X and Y and functions as a mediator between X and Y, we can infer that the relationship between X and Y will be strong when it is supported by Z. This reasoning leads to Hypothesis 3.

Hypothesis 3: A tie with an acquaintance formed through an introduction by another researcher is stronger than a tie with an acquaintance from an academic conference.

It is a feature of industry–academia ties that intermediary organisations often involve. Industry–academia mediating offices not only connect academic and corporate researchers but also support and manage their joint efforts towards patent applications and entrepreneurship (Lee, 2014). Therefore, as long as a project continues, the industry–academia mediating office will support the maintenance of strong ties. Thus, we can expect two researchers who first met through an industry–academia mediating office to form a stronger tie for a personal network than two researchers who first met at an academic conference with no intermediary. However, whether a strong tie will continue after the project ends seems to depend on how academic and corporate researchers build trust during the project period. Based on the above considerations, Hypothesis 4 is derived.

Hypothesis 4: A tie with an acquaintance formed through a mediating organisation is stronger than a tie with an acquaintance from an academic conference.

Method

This study analyses data collected in 2018 by “A Survey on Researchers’ Knowledge/Information Exchange and Dissemination” which targets researchers affiliated with national universities or institutions in Japan; surveys are better suited for collecting large samples than are interviews. Japan has 47 prefectures, each with at least one national university. First, we selected one or two national universities for each prefecture. Next, a total of 2,060 researchers were randomly chosen from the selected universities according to the following two criteria: participants in each prefecture included both life science/medical science experts and natural science/engineering experts; seven large universities had to be represented by 60 participants each, and all other universities had to be represented by 40 participants each to conduct a survey of approximately 2,000 people with budget constraints. A database provided by the Japan Association of National Universities was used for the sampling. A total of 495 valid responses were returned by post, for a 24.9% effective response rate after 33 pieces of undelivered mail were subtracted from the total.

In addition, the same questionnaire was sent to researchers at all three Japanese national research institutes which are designated as the National Research and Development Agencies (NRDA), the resources and operations of which are governed by special law. National Laboratory A is a multidisciplinary laboratory that covers many fields, including physics, engineering, chemistry, computer science, biology, and medical science. National laboratory B has developed a wide range of industrial technologies. National Laboratory C specialises in research on substances and materials. In accordance with the size of each laboratory, 940 researchers from A, 620 researchers from B, and 440 researchers from C were randomly selected from the database of researchers provided by each institute. A questionnaire was mailed to these 2,000 researchers, and 509 valid responses were collected; after 18 undelivered questionnaires were subtracted from the total, the effective response rate was 25.7%.

In total, 1004 valid responses were collected from national universities and national research institutes. The response rate, which was approximately 25%, is relatively high for a survey targeting individuals. If the sheet had been delivered to each researcher by his or her organisation, the response rate would have been higher. However, because the questionnaire included questions about the participants’ personal behaviour, the involvement of organisations was undesirable. Only those who were willing to answer the questions participated in the survey.

The model used to test the hypotheses is as follows :

Note that TIE represents the strength of the tie of academic researcher i with the corporate researcher with whom i most frequently exchanged knowledge for the past year, TRIGGER is the vector of triggers for tie formation, CONTROL is the vector of the control variables, and $\delta$ is an error term.

The details of the variables are shown in Table 1. The strength of ties, which is the dependent variable of Eq. (1), was measured by the frequency of knowledge exchange with a partner over the past year (short-term tie strength) and the duration of knowledge exchange (long-term tie strength). The former is the frequency variable, and the latter is the duration variable. In the survey, respondents were asked to select one of five options concerning the frequency of knowledge exchange (once a year, two to three times a year, once a month, once a week, or two to three times a week). Given that the mode of frequency was two to three times a year, those who selected a frequency of once a month or more were considered to form a strong tie, and logistic regression analysis was employed for the model with the frequency variable. Long-term tie strength was measured by the log-transformed value of the number of years of knowledge exchange with the partners up to the time of the survey. Multiple regression analysis was applied to the model with the duration variable.

TRIGGER represents four variables, with “academic conferences” being the reference category: university, past work, peer introduction, and mediating organisation. The “university” trigger includes relationships with respondent’s senior, junior, friend and professor. The “past work” trigger includes relationships with a former colleague and a student. The peer introduction and the mediating organisation variables mean introduction by a peer researcher and an intermediary organisation, respectively. Sixteen respondents who chose the other option for trigger and 46 respondents who made multiple choices were excluded from the target of this study. There were 508 respondents with at least one knowledge exchange partner in the industry who answered all the questions related to the variables required for the analysis and were not subject to the above exclusions. We used these samples to test Hypotheses 1–4.

The ability variable is included in the control variables. Studies of academic engagement have shown that highly capable researchers engage in it (summarised in Perkman et al., 2013). Academic engagement is defined as knowledge-related interactions between academic researchers and non-academic organisations, with the exceptions of teaching and commercialisation (Perkman et al., 2013). It includes collaborative research, contract research, consulting, and informal networks with practitioners. Therefore, highly capable researchers are expected to form strong ties across industry–academia boundaries. The acquisition of external knowledge is part of absorptive capacity. Cohen and Levinthal (1990) regarded absorptive capacity as a function of the level of knowledge already possessed, which includes basic skills, shared language, and knowledge of recent scientific and technological developments in a given field. However, as Zahra and George (2002) and Michailova and Mustaffa (2012) noted, subsequent empirical studies have used various measures for absorptive capacity.

Nonetheless, communication skills and expertise are the measurements used in many studies, such as Szulanski (1996) and Matusik and Heeley (2005). Therefore, this study used English skills, which are common means of communication among researchers and expertise in their research fields, to measure ability. The statements in the questionnaire used to measure these abilities were “I am confident in my English communication skills” and “I have more expertise than the average R&D personnel in my field”. With respect to the basic skills of Cohen and Levinthal (1990), we employed two measurements: “I have a great ability to explain and make others understand” and “I have a great ability to acquire and understand others’ knowledge”. The former is persuasive skills, and the latter is learning ability, two of which Sun and Scott (2009) identified barriers to knowledge transfer. Principal component analysis was performed using the answers measured on a 5-point Likert scale (does not apply = 1, applies = 5) for each of the above four items, and one principal component with eigenvalues greater than 1 was obtained (see Table A.1). Therefore, the principal component scores were used as the values for the ability variable.

If the number of papers published or patents filed is used as data representing the ability variable, the opposite causal relationship in which researchers produce many papers and patents because they have a strong tie is possible, which is the problem of endogeneity. Therefore, this study used items in the questionnaire survey to measure ability, which were evaluated by the respondents themselves. Because this study employs data obtained from the same questionnaire for both the independent and dependent variables, Common Method Bias (CMB) may influence the results. However, because the results of Herman’s single factor test show that the total variance for a single factor is approximately 10%, we can conclude that CMB does not affect our data.

Results

The basic statistics of each variable in Table 1 are shown in Table 2. The average values of the four variables regarding triggers for network formation show that “university” accounts for 17%, “past work” for 8%, “peer’s introduction” for 22%, and “mediating organisation” for 17%; thus, the remaining 36% is “academic conference”, which is the reference category of our model. In other words, academic conferences provide important opportunities for academic and corporate researchers to meet. Regarding the dependent variables, the average frequency shows that approximately one-third of the researchers have a strong tie. The average duration variable is 1.53, which means that the average period of knowledge exchange is approximately 4.6 years. While the correlation matrix is not presented here owing to space limitations, the highest correlation between two independent variables is 0.263 for the age variable and the tenure variable. All other correlations are less than 0.260.

The results of the analysis of the model are shown in Tables 3 and 4. According to Table 3, which shows the results of the model with tie strength measured by the frequency of knowledge exchange, the two trigger variables are significantly positive: mediating organisation at the 5% level and peer introduction at the 10% level. A strong tie is more likely to be formed with a researcher introduced by a mediating office or a researcher introduced by a peer researcher than with an acquaintance from an academic conference. However, the coefficients of the university variable and the past work variable are positive but not significant. Therefore, Hypotheses 3 and 4 were supported, and Hypotheses 1 and 2 were rejected. For the marginal effects, the largest effect of 1.959 is observed for the mediating organisation variable. In other words, academic and corporate researchers who meet through an introduction by an organisation such as an industry–academia mediating office are the most likely to form strong ties, as measured by the frequency of knowledge exchange, which is approximately twice as high as when they meet at academic conferences. This may be because a mediating organisation is not merely a referral but also supports collaboration throughout the time when the joint research is conducted.

Table 4 shows the results of the model with the duration variable being the dependent variable. The university variable and the past work variable are significantly positive at the 1% and 5% levels, respectively. In other words, the ties formed in universities or workplaces are likely to become strong ties that last for the long term. Therefore, Hypotheses 1 and 2 are supported. School-aged ties or those formed through past work are rooted in the experience of sharing “Ba” and activities. Thus, they seem to be bound by affect-based trust. The coefficient of the peer introduction variable is positive but not significant, which contradicts Hypothesis 3 when the strength of ties is measured by the duration of knowledge exchange. The mediating organisation variable is noteworthy. It shows a significantly negative coefficient at the 1% level, which is the opposite of the sign of the coefficient obtained when tie strength is measured by the frequency of knowledge exchange. In other words, mediating organisations function as facilitators of knowledge exchange between academic and corporate researchers in a project period. However, they do not seem to help develop affect-based trust, which is key for knowledge exchange to continue after the project is over.

In summary, Hypotheses 3 and 4 are supported for short-term tie strength, and Hypotheses 1 and 2 are supported for long-term tie strength. Notably, for the mediating organisation variable, the sign of the coefficient changes depending on whether tie strength is measured by the frequency of knowledge exchange or years of knowledge exchange. Hypothesis 4 is supported in the short term but rejected in the long term.

Among the control variables, only the age and tenure variables have significantly positive coefficients in both Tables 3 and 4. Thus, I infer that experienced researchers with tenure are more likely to overcome the institutional distance between industry and academia. This may suggest that some experience is needed to connect Mode 1 and Mode 2 knowledge.

Discussion

This study revealed that ties formed through peer researchers’ introductions and mediating organisations were stronger than those formed at academic conferences in the short run, whereas ties formed at schools or workplaces were stronger than those formed at academic conferences in the long run. In other words, a strong tie in the short run is a tie formed by a third party’s intervention, whereas a strong tie in the long run is a tie developed over a relatively long period, sharing “Ba”. This difference in the impact of the triggers for tie formation on tie strength between the short and long term may be related to McAllister’s (1995) discussion of cognition-based and affect-based trust: cognition-based trust is more superficial and less special than affect-based trust is, and affect-based trust develops on the basis of cognition-based trust. Short-term strong ties can be formed when an intermediary ensures the cognitive level of the partner. However, for this to develop into affect-based trust, interactions over time seem necessary. Ties with former colleagues and peers of school age are likely to fulfil this condition and become strong in the long run.

Xu et al. (2019) argued that three types of personal networks differ in terms of triggers for tie formation, knowledge function, and geographic scope on the basis of interviews with entrepreneurs and senior managers of biotechnology firms in China. First, bonding personal networks are formed on the basis of private social relations that are deeply rooted in common experiences. Family members, relatives, and middle school or university schoolmates form bonding networks. Second, bridging personal networks are formed on the basis of professional relationships and careers. Teachers, former bosses and colleagues, project members, or close business partners form bridging networks. Third, linking personal networks refers to “the socially thinnest, culturally neutral, and specific matter-oriented relations” (p. 6). They connect people known through online technical communities, professional meetings, trade fairs, etc. The authors found that the degree of social embeddedness decreased in the order of bonding, bridging, and linking networks. In other words, the ties in the networks could be weaker in this order. According to their classification, ties with university schoolmates are included in the bonding network, whereas ties with people met at academic conferences are categorised as the linking network with the lowest embeddedness level. My finding that “a tie with a university acquaintance is stronger in the long term than a tie with an acquaintance from an academic conference” is consistent with the findings of Xu et al. (2019). Similarly, their assertion that networks with former colleagues categorised as bridging networks have higher embeddedness levels than ties with people met at academic conferences, which are included in linking networks, is consistent with my finding that “A tie with an acquaintance from past work is stronger in the long term than a tie with an acquaintance from an academic conference”.

The results of this study also recall the argument of Juhász and Lengyel (2018) that factors for the creation of knowledge networks differ from those for the persistence of knowledge networks in clusters. Their study, which targets Hungarian interfirm networks, not personal networks, finds that triad closure is a factor in the creation of ties but not in their persistence. In other words, the presence of a third party contributes to the formation of ties but not to their maintenance. This is similar to the finding of my study that ties formed through the intervention of peer researchers and mediating organisations are stronger in the short term but not in the long term. Similarly, the findings of my study correspond to Staber’s (2011) findings. The author analysed networks of small business owners in the German textile and clothing sector, and demonstrated that social ties created through third-party referrals are less durable than ties created through direct relations.

As noted above, we can observe some common factors for tie formation and tie strength that are applicable to both personal networks among academic and corporate researchers, i.e., the target of our study, and other networks, which are of interest to prior studies. Nevertheless, one of the characteristics of personal networks among academic and corporate researchers is the contribution of mediating organisations. Industry–academia collaboration has been promoted by policy, and collaborative research and other joint activities have been realised through the assistance of such organisations. Notably, the ties formed through mediating organisations are strong in the short term but weak in the long term. Short-term strength might be due to the organisation’s support. The long-term weakness is probably because the presence of an intermediary may weaken the relationship between two actors and make it difficult for cognition-based trust to develop into affect-based trust. Another possible reason is that industry and academia originally differ in terms of knowledge of interest, as mentioned in the second section: academic (corporate) researchers are interested in Mode 1 (2) knowledge. Therefore, there are only limited situations in which academic researchers need knowledge that is provided only by corporate researchers, and vice versa. In other words, the lack of long-term complementary knowledge, which is a driver of industry–academia collaboration, is another possible reason for the weakness in the long run.

On the basis of these findings, this study can provide suggestions for management and public policy so that organisations can absorb external knowledge through individuals’ personal ties. First, academic conferences provide important opportunities for the formation of knowledge exchange networks among academic and corporate researchers, who rarely meet each other, although ties formed through this trigger are not strong, both in the short and long term. Participation in academic conferences involves time costs outside normal work as well as financial costs such as transportation and conference registration fees. Therefore, management that encourages researchers to attend conferences by allowing time for participation or by covering financial costs may help lower the connection barrier between academic and corporate researchers.

Second, given that ties formed through past work experience are strong, one way to absorb external knowledge through personal networks is to hire people who have work experience rather than new graduates. In particular, people have work experience in academia (industry) if the employer is a firm (university). The Japanese labour market is relatively immobile due to long-term employment practices (Kitagawa et al., 2018); in particular, mobility from academia to industry is very limited (NISTEP, 2023). Policies that promote mobility, including a cross-appointment system, also contribute to the formation of strong ties in personal networks.

Third, given that introduction by peer researchers leads to strong ties in the short term, managers are instructed to encourage employees to be embedded in a close network that includes researchers outside the organisation. Although effective management which is specific for knowledge sharing with people outside the organisation has rarely been explored, the findings on knowledge sharing in general in previous studies may be effective, to some extent, for knowledge sharing through personal networks connecting academic and corporate researchers; knowledge sharing requires not only the ability and motivation of knowledge holders but also the environment, which provides them with time, funding, materials, autonomy, opportunities, and support from supervisors and peers for knowledge sharing (Siemsen et al., 2008; Minbaeva, 2013; Andreeva and Sergeeva, 2016; Llopis and Foss, 2016; Foss et al., 2009; Cabrera and Cabrera, 2005; Cabrera et al., 2006). Conversely, management that strongly controls employees with competition clauses may result in fewer informal links to the outside (Dahl and Pedersen, 2005).

Fourth, the results of this study indicate that specialised offices for industry–academia collaboration contribute to the formation of strong ties in the short term. However, the ties mediated by such an organisation tend to be weak in the long term. If mediating organisations connecting universities and industry recognise and support the use of ties formed through school-age or past work, they will contribute to the formation of strong ties not only in the short run but also in the long run.

If we assume that personal networks of knowledge exchange are more likely to form within a region/cluster because greater geographic distance limits opportunities for personal interactions (Ponds et al., 2010; Hoekman et al., 2010; Katz, 1994), public policies to create regional clusters and corporate measures to locate their headquarters or branch office near the centre of excellence or in the cluster are the best to absorb knowledge through personal networks between academic and corporate researchers. However, the results of this study suggest that various other policies and management measures can also promote knowledge absorption through personal networks between academic and corporate researchers.

Conclusion

According to previous studies, strong ties contribute to innovation. However, not only are the knowledge exchange networks among academic and corporate researchers small, but the ties are also likely to be weak compared with the ties connecting academic researchers. Therefore, this study analysed the strength of ties among academic and corporate researchers in terms of triggers for tie formation to clarify how strong ties are formed and to consider policies and management to increase the absorption of external knowledge. First, we consider the five types of triggers for network formation and then the impact of these triggers on the strength of ties that connect Japanese researchers at national universities and public research institutes with corporate researchers. The empirical part of this study revealed that the five types of triggers for network formation differently affect short-term tie strength, measured by the frequency of knowledge exchange in the past year, and long-term tie strength, measured by the number of years of knowledge exchange. From a short-term point of view, ties formed through the intervention of a third party, such as a peer researcher or a mediating organisation, are stronger than those formed through a direct encounter at academic conferences. On the other hand, with respect to long-term tie strength, ties formed through shared experience at school or the workplace are stronger than those formed through encounters at academic conferences.

Trust, which is a driver of knowledge exchange between industry and academia, is at the core of the mechanism through which the triggers for tie formation are related to tie strength. Depending on these triggers, the length of time in which the parties share a “Ba” and can interact/communicate is different. This might influence the development of affect-based trust. In addition, depending on the triggers for tie formation, the involvement of a third party differs. The third party ensures the partner’s competence and accountability. This supportive evidence leads to cognition-based trust. However, for cognition-based trust to develop into affect-based trust, repeated interactions and communication over time might be needed.

Prior studies on the absorption of external knowledge through personal networks have focused on the types of knowledge acquired (Huber, 2013; Dahl and Pedersen, 2004; Xu et al., 2019), the impact on innovation (Fitjar and Huber, 2015; Xu et al., 2019), factors that increase informal contacts (Dahl and Pedersen, 2005), the geographic distance of informal contacts (Huber, 2012b; Fitjar and Huber, 2015; Xu et al., 2019), and the context of building and maintaining personal networks (Huber, 2012a). They have made important discoveries but have not proved the mechanisms by which the discoveries have occurred. The first academic contribution of this study is its focus on the strength of ties in personal networks between academic and corporate researchers, i.e., a subject that previous studies have not addressed. The second academic contribution is to show the main mechanism linking triggers for tie formation and the strength of ties, i.e., cognition-based trust and affect-based trust, and to prove the hypotheses that are theoretically derived from this mechanism. This study also makes a practical contribution: suggestions for management and public policies for organisations to absorb external knowledge through personal networks between academic and corporate researchers.

This study has the abovementioned contributions but also has some limitations that are potential avenues for future research. First, the sample for this study included academics from national universities and research institutes and did not include information from their partners in industry. As knowledge exchange is a two-way activity, the frequency and number of years of exchange may not differ when corporate researchers are asked, but the environment in which corporate researchers form and maintain networks with outside researchers may differ from academic researchers’ environment because of the aforementioned institutional differences. Therefore, conducting a similar study with a sample of corporate researchers is a topic of future research.

Second, this study analysed only data from Japanese researchers. As the history and policy measures used to promote industry–academia collaboration vary from country to country, the findings of this study regarding the strength of ties formed through the help of mediating organisations may not be found in other countries. In addition, Japan is characterised by long-term employment practices and relatively immobile labour markets. In particular, the mobility of human resources from academia to industry is rare. This may affect the degree of development of industry–academia networks and the factors involved in the formation and maintenance of industry–academia networks. Therefore, a future study should conduct another analysis using data from other countries and international comparisons.

Third, because the data used in this study are cross-sectional, they may contain selection bias and only include information on maintained ties, not extinct ties. It is a future task to analyse when and how ties formed through “introduction by peer researchers”, “acquaintances from academic conferences”, and “introduction by mediating organisations” weaken or are severed and how they vary depending on the trigger of tie formation.

Fourth, the findings indicate that ties formed through mediating organisations are strong in the short run but weak in the long run. However, previous studies have discussed various aspects of the performance and efficiency of boundary-spanning organisations and factors that improve or worsen their performance via objective or subjective measures. Factors such as organisational design and structure, human capital (experience, expertise, and skill sets of personnel), relationships with researchers in both industry and academia, group economic incentives (financial situation, internal/external support for the organisation) and university intellectual property strategies have been discussed and examined (Siegel et al., 2003; Pronay et al., 2021; Faccin et al., 2022; Zmuidzinaite et al., 2021). Therefore, future research could analyse which industry–academia mediating offices contribute to the formation of strong ties in the long run.

Acknowledgments

I would like to thank anonymous referees of this journal. This work was supported by a Grant-in Aid for Scientific Research (No. 16K03714) from the Japan Society for the Promotion of Science.

Appendix A

ORCID

Yukiko Murakmi  https://orcid.org/0000-0002-6102-1193