Introduction

Internationalization is inherent to higher education in the 21st century (Baruffaldi and Landoni 2012; Audretsch et al. 2015). Since the 1990s, academics increasingly move across international borders for different durations and reasons (Rostan and Höhle 2014; Teichler 2015). Many countries have developed policies and programs to support international mobility by students and faculty (Scellato et al. 2015). For example, in Europe, the Erasmus Program and the European Research Area increase the opportunity for academics to be internationally mobile (Maggioni and Uberti 2009). The United States has its long-established J and H visas for short-term academic exchanges (Wang et al. 2019). The China Scholarship Council (CSC) and Horizon 2020 projects provide financial assistance to Chinese academics who wish to study abroad (De Moortel and Crispeels 2018). The idea behind these mobility programs is that the accumulation of networks and collaborations abroad enhances the academic’s individual capabilities, productivity and career development, and that these benefits extend to a regional or country level in terms of increased welfare through overall knowledge exchanges and enlargement of the regional or national knowledge stock (Bauder 2020; Veugelers and Van Bouwel 2015; Regets 2007).

Consequently, a scholarly debate emerged about the impact of international mobility on academics (Teichler 2015). For example, when it comes to international student mobility, we expect increases in international competences or in comparative reasoning, and, when it comes to staff mobility, increases in scientific output (e.g. publications, citations, or patents) or academic quality. This debate has resulted in a number of scholars (e.g. Wang et al. 2019; Yasuda 2016; Krabel et al. 2012; Edler et al. 2011; Cañibano et al. 2011; Stephan and Levin 2001), although using different operationalizations and conceptualizations, who study the relationship between international academic mobility and knowledge transfer.Footnote 1 Within this debate, we aim to untangle the relationship between international academic mobility and academic entrepreneurship further. Academic entrepreneurship refers to the efforts undertaken by an academic to commercialize an invention, including patenting and licensing, and the creation of a start-up company (Siegel and Wright 2015). These efforts constitute the more formal knowledge transfer activities. In their work, Siegel and Wright (2015) note that mobility is increasingly considered a way to foster academic entrepreneurship, next to the existence of technology transfer offices and science parks. This notion draws on the assumption that knowledge and/or interpersonal networks gained through international mobility are conducive to the academics’ engagement in the entrepreneurial activities (Krabel et al. 2012). However, to the best of our knowledge, no study has actually tested this strong assumption.

To address this gap, we study the relationship between international mobility and entrepreneurial knowledge from the individual academic perspective. We choose to focus on temporary international mobility, which holds the academic’s intention to return to the home country. Temporary international mobility is an increasingly relevant phenomenon in academia, especially amongst PhD students and post-docs (Teichler 2015; Edler et al. 2011), while permanent mobility of academics is rather limited in practice (Cervantes and Guellec 2002). Our premise is that international temporary academic mobility is conducive to the academic’s entrepreneurial knowledge and that interpersonal social networks mediate the relationship. In doing so, we add insights to the impact of international temporary mobility and interpersonal social networks on the accumulation of entrepreneurial knowledge of academics (Cañibano et al. 2008); an outcome which is often explicitly or implicitly assumed.

Our study provides support to the idea that academics reap benefits from their internationalization efforts, adhering to the brain gain discourse. We also increase our understanding on the relationship between of international mobility and academic entrepreneurship, as different authors use different operationalizations without further clarification. It is striking to see how, although international academic mobility has been emphasized for several decades, our understanding on the phenomenon and its impact remains scattered, underdeveloped and mis-aligned (Veugelers and Van Bouwel 2015; Teichler 2015).

Literature Review and Hypotheses Development

International Temporary Academic Mobility

Academic mobility refers to physical – as opposed to virtual - movements of students and faculty across borders (Rostan and Höhle 2014). Geographically, academics can move across national borders, i.e. international academic mobility, and be mobile domestically, i.e. domestic academic mobility. When it comes to studying the impact of international academic mobility on knowledge transfer or academic entrepreneurship, we agree with the view of Edler et al. (2011) that one should differentiate between temporary and permanent movements of academics. While temporary mobility refers to academics going abroad to gain experiences and come back to the home country, permanent mobility does not hold the particular intention to come back. Such distinction is crucial to debates on brain drain and brain gain. While brain drain refers to international mobility negatively affecting the academic’s home country, e.g. through decreases of technological capabilities or overall competitiveness (Adams 1968), brain gain puts emphasis on resulting benefits for the home country, e.g. increases in welfare through overall knowledge exchanges and better career development of individual academics (Regets 2007).

Literature on permanent mobility mainly devotes attention to the (return) migration of academics (e.g. Wang et al. 2019; Gibson and McKenzie 2014). Davenport (2004) and Trippl (2013) provide support to the idea that internationally mobile academics return home with cutting-edge knowledge and networks and act as important transmitters of technology and tacit knowledge. Other scholars (e.g. Krabel et al. 2012; Stephan and Levin 2001) find that entrepreneurial exploitation, operationalized as nascent entrepreneurship or the creation of spin-off companies, is stimulated when the academic is foreign born or educated. Similarly, scholars find a positive impact on entrepreneurial opportunity identification through academic patenting or find a trend in the inventive activities of foreign-born academics (e.g. Hunt and Gauthier-Loiselle 2010; Wadhwa et al. 2008). Jonkers and Tijssen (2008) and Jonkers and Cruz-Castro (2013) find a positive impact of permanent international mobility on academic publications.

Temporary international academic mobility is an increasingly relevant yet often underestimated phenomenon (Teichler 2015). The Changing Academic Profession (CAP) survey, for example, concluded that one third of all academics surveyed engaged in some sort of temporary international mobility. The distinction between the different types of temporary mobility is subtle and debatable, especially when it comes to defining the duration of the activities (Table 1). Arguably, such definition is arbitrary anyway (Cañibano et al. 2011). For example, an academic might be participating to workshops and conferences while on a research stay abroad or a particular teaching activity abroad can take only a few days, several weeks, or can happen on a recurrent basis. In such cases, it becomes very hard to assess the impact of (isolated) experiences on one’s knowledge or networks gained.

Table 1 International academic mobility: an overview of (some) distinctions made
Full size table

Despite the unclear boundaries, scholars have started to see value in the role of non-permanent movements of academics in knowledge transfer and networking activities (Cañibano et al. 2011; Edler et al. 2011). Different temporary international mobility experiences shape one’s knowledge and networks differently. For example, Teichler (2015) notes that there is a distinction between students who study abroad for only a semester and those who participate to a full study program abroad, since the former imposes contrasting learning environments at different universities impacting one’s knowledge differently. Similarly, short-term international experiences may impact knowledge transfer activities differently and play a different role in the development of international connections (Edler et al. 2011; Melkers and Kiopa 2010). Melkers and Kiopa (2010), for example, indicate that short-term visits are especially useful in the consolidation and maintenance of international networks. Edler et al. (2011), who focus on intra-sectoral research mobility, find that temporary international mobility is conducive to the academic’s overall collaborations with industry.

International temporary academic mobility thus comprises a range of inter- and intra-sectoral experiences: university, industry, or governmental visits, participation in international conferences and workshops, research stays abroad, summer schools, international project meetings, contract research abroad, studies abroad and so forth. These experiences range from short-term moves of a few days/weeks to longer movements over one year and form by no means a coherent bundle of activities. In this study, we choose to work with the full spectrum of temporary international mobility experiences.

Entrepreneurial Knowledge

To understand the relationship between international academic mobility and entrepreneurial knowledge, one should especially consider the perspective of the individual academic and the factors that drive the engagement in entrepreneurship (Goethner et al. 2012; Eckhardt and Shane 2003). Scholars acknowledge that, in entrepreneurship, knowledge is a key resource (Hayter 2016). In this respect, the knowledge-based view states that knowledge accumulation over time creates a knowledge base that allows someone to identify certain entrepreneurial opportunities (Venkataraman 1997; Alvarez and Busenitz 2001). People possess different knowledge bases which change over time as knowledge is accumulated through different life experiences (Shane 2002). Mobility is an integral part of the academic’s life experiences.

Thus, for an entrepreneurial opportunity to be recognized, an academic needs insights which appear through the availability of accumulated knowledge within one’s knowledge base (Jacobs 1969). Heterogeneity is identified as a key attribute to the development of such insights (Alvarez and Busenitz 2001). Especially the availability and increases of heterogeneous knowledge positively affects one’s ability to recognize entrepreneurial opportunities since it allows academics to be more aware of potential uses, applications, and to assess commercial value (D’Este et al. 2012; Bercovitz and Feldman 2008; Alvarez and Busenitz 2001). Four knowledge categories are relevant in the accumulation of entrepreneurial knowledge. Market knowledge points to the need to understand markets, ways to serve markets, and customer problems (Shane 2002). Widding (2005) synthesizes the literature on the business knowledge needed to start a company and identifies three additional categories: product knowledge, organizational knowledge, and financial knowledge. Product knowledge refers to technological knowledge, production knowledge, and knowledge related to service offerings. Organizational knowledge refers to knowledge on (human resource) management and organizational structures. Finally, financial knowledge refers to knowledge on funding, financial management and taxes. We thus consider entrepreneurial knowledge as a compound construct of heterogeneous knowledge components on how to start and operate a business including know-how on opportunity identification or exploitation and on functional aspects of starting and running a business (Honig 2004; Pretorius et al. 2005).

This work poses that international temporary mobility positively contributes to one’s entrepreneurial knowledge. Edler et al. (2011), for example, argue that the duration and frequency of international visits positively influence the academics’ knowledge base. Academic mobility leads to a greater ability to find and develop rare or unique skill sets, which leads to cognitive integrations and the facilitation of opportunity identification (Olmos-Peñuela et al. 2015). Also, the exposure to specific and different country and/or institutional settings highly influences one’s knowledge base (Busenitz et al. 2000). Indeed, internationally mobile academics are exposed to knowledge they did not have access to before (Edler et al. 2011). This leads these academics to hold a greater heterogeneity of ideas, perspectives, assumptions and creative techniques than those who did not have similar experiences before (McEvily and Zaheer 1999). Thus, through exposure to international experiences, academics accumulate external knowledge, including entrepreneurial knowledge, over time which augments their knowledge bases (Cañibano et al. 2008; Politis 2008). We hypothesize that:

Hypothesis 1

International temporary academic mobility positively influences the academic’s entrepreneurial knowledge.

The Mediating Role of Social Networks

The role of social networks in international mobility has been recognized by many scholars (Cañibano et al. 2008). International mobility creates and shapes social networks (Scellato et al. 2015). At an interpersonal level, a social network is defined as a set of individuals and a set of linkages between these individuals (Brass 1992). Several macro-level concepts, like Mode 2, Triple Helix, and Post-Academic Science point to increased importance of social networks in academia (van Rijnsoever et al. 2008). Academics are involved in different network types, e.g. university networks (contacts within one’s university), external university networks (contacts with other university researchers), and industrial networks (contacts with private companies). These networks may be of personal (e.g. friends or colleagues) or professional (e.g. mentors or business contacts) nature (Fernández-Pérez et al. 2015). Travels abroad allow for face-to-face meetings and interactions (Urry 2002; Bienkowska et al. 2011) and bring along investments of effort, time and money, which result in international interpersonal relationships and networks (Bienkowska and Klofsten 2012; Orazbayev 2017). Through social networks, international mobility then influences individual knowledge (Cañibano et al. 2008). We model this causal relationship in Fig. 1. International temporary mobility causes changes in the academics’ interpersonal network, which in turn causes changes in the academic’s entrepreneurial knowledge (Wu and Zumbo 2008). We thus argue that one’s interpersonal network explains the process of how temporary international mobility impacts entrepreneurial knowledge; providing a deeper understanding of the relationship (Baron and Kenny 1986). We investigate this impact through the strength and the size of interpersonal social networks, two properties that typically influence the transfer of information and knowledge (Fuentes-Fuentes et al. 2010). Network strength constitutes a relational property, while size represents a structural property (Phelps et al. 2012).

Fig. 1
figure 1

Conceptual framework on international temporary mobility, interpersonal social networks, and entrepreneurial knowledge (source: authors)

Full size image

In terms of network strength, personal relations can be classified as either formal or informal according to the weak or strong tie that binds the individuals (Granovetter 1973). Weak ties provide access to new information by means of bridging disconnected individuals, while strong ties are more likely to provide redundant information. Still, some controversy exists on the effect of formal or informal networks on entrepreneurship and the usefulness of knowledge transferred. Cetin et al. (2016), for example, find that formal networks have a negative effect on entrepreneurship, while informal ones have a positive effect. Other studies (e.g. Davidsson and Honig 2003; Casson and Giusta 2007) show that formal networks also provide business opportunities and deliver valuable information. Additionally, literature notes that informal relations are useful to transfer tacit and heterogeneous knowledge and that formal networks help to transfer explicit and homogenous knowledge (Villanueva-Felez et al. 2013; Zaheer et al. 2010; Levin and Cross 2004). Taken together, these studies support the idea that social networks gained, in our case through international temporary mobility, positively contribute to one’s entrepreneurial knowledge and that the strength of one’s social network plays a role in determining the usefulness of the knowledge received. We thus hypothesize that:

Hypothesis 2a

The relationship between the academic’s temporary international mobility and entrepreneurial knowledge is mediated by one’s interpersonal social network strength.

In terms of network size, literature seems to agree on the fact that every relationship is able to provide information which turns into useful and meaningful knowledge when given context and interpretation by an individual (Chou 2005; Levin and Cross 2004; Granovetter 1973). From a knowledge-based view, academics receive and assimilate new knowledge through every network tie – whether formal or informal (van Rijnsoever et al. 2008; Hoang and Antoncic 2003; Rasmussen et al. 2015). Thus, international mobility increases the size of the academic’s interpersonal network. The increase in network size extends one’s knowledge base through the accumulation of multi-faceted heterogeneous entrepreneurial knowledge components (Edler et al. 2011) (Fig.1). We hypothesize that:

Hypothesis 2b

The relationship between the academic’s temporary international mobility and entrepreneurial knowledge is mediated by one’s interpersonal social network size.

Methodology and Data

Methodological Approach

We perform a structural equation model (SEM) to test our hypotheses. A SEM has become a dominant analytical tool to test mediated models through the use of path analysis and latent variables (Hair et al. 2014). Path analysis is a form or multiple regression statistical analysis that allows to investigate patterns of effect within a model. Latent variables, which we will also refer to as constructs, are underlying variables that cannot be observed directly but consists of one of more observable indicators (Wong 2013). We use SmartPLS 3 software (Ringle et al. 2015) to run a Partial Least Squares (PLS) – SEM, which provides iterations to maximize the explained variance of endogenous constructs (Wold 1974).

Research Context

The Chinese setting is particularly relevant to our study. International mobility has intensified over the past years through supporting programs, such as the Ten Thousand Talents Program, Erasmus programs, China Scholarship Council grants, and Horizon 2020 projects (Wang et al. 2019). As such, Chinese universities increasingly collaborate on student or staff exchange programs and on joint PhDs. As an indication, the number of returned study abroad students and staff increased by 2179% in the period 2002–2015 (Li 2017). Talent pooling and training stimulates knowledge generation and knowledge transfer. International PhD programs or joint graduate schools allow for cross-country research and access to technologies that otherwise would have remained unattainable. These incentives reflect in the amount of university encouragement and support Chinese academics perceive with regards to international mobility; 6.4% of the Chinese academics perceived no/low university encouragement and 13.3% indicated that no/little university support system was in place with regards to international mobility (own data).

Within Chinese society and culture, interpersonal networks play an important role. China embraces the concept of quanxi, the fundamental web of Chinese interpersonal relations, which is based on implicit mutual interest and benefit (Xin and Pearce 1996; Buckley et al. 2004). Xin and Pearce (1996) find that such relations even substitute more formal arrangements in a business context and that they are fertile ground for the transfer of inter-organizational knowledge. A specific stream of literature devotes attention to the role of guanxi in knowledge transfer (e.g. Davison et al. 2018; Ramasamy et al. 2006; Buckley et al. 2004). While such in-depth research is out of the scope of this study, we do note its importance with respect to the Chinese context and the mobility of Chinese academics as it characterizes our research sample as one that pays attention to the development of interpersonal networks.

Data and Research Sample

Although the internationalization of academia has received significant scholarly attention over the past decades, the quality and availability of data on international students and staff is worrisome (Teichler 2015). This is especially true for data on temporary mobility. EUROSTAT and the OECD, for example, recommended not to include student mobility shorter than one year. Teichler (2015) argues that surveys provide the most promising tool to increase our understanding on international mobility and its impact.

We launched an online survey amongst Chinese academics running from April till August 2019. The survey was translated using a double-translation design (English being the original language) and validated, linguistically and culturally, to fit the Chinese context. The survey was constructed and disseminated through Qualtrics Software (www.qualtrics.com). Academics were contacted through their publicly available email addresses which we retrieved from the websites of the Chinese Academy of Sciences (http://www.cas.cn/) and DaoshiFootnote 2 (https://daoshi.eol.cn). We sent two reminders, with one week in-between mailings. Our survey was sent to 22,285 academics of which 1,981 (or 9%) engaged in the survey. However, only 319 academics provided us with fully completed responses. We believe that this discrepancy is largely due to the survey set-up. The average completion time (12 minutes and 12 seconds) exceeded the recommended threshold of nine minutes and a considerable amount of questions was asked in matrix form. The combination of these two elements lowers completion rates drastically (www.qualitrics.com). In addition to this, the recent cybersecurity law in China (2016) may hamper the willingness of respondents to disclose personal data to Western platforms (Mei and Brown 2018). From the usable responses, we excluded 38 cases which did not report any international activities and/or did not have the Chinese nationality. In line with our operationalization of the international temporary academic mobility construct, we also excluded all cases that reported any permanent mobility activities. The above procedure leaves us with 281 usable responses and a sample error of 4.9% at the 90% confidence level (Z = 1.65, p = q = 0.50).Footnote 3

As PLS-SEM is even capable to handle sample sizes smaller than 200, our sample is also sufficiently large to carry out any analyses (Hoyle 1995). Table 2 provides an overview of our sample characteristics. In terms of gender, the proportion of males is in line with the proportion reported in other survey studies (e.g. 63% in Shen (2008) and 54% in Stanfield and Shimmi (2014)) and in line with Chinese academic staff population predominantly being male.Footnote 4 In terms of the research area, our sample is mainly represented by academics from the natural sciences and technological disciplines (Morhman et al. 2011). This results from our survey partly being directed towards academics from universities which are part of the Chinese Academy of Sciences. Our sample tends to lean towards academics holding a professorship (74%). In Shen’s (2008) study, which covers a much larger sample, this proportion only represents 57%. We decided to retain the responses of a variety of academic positions in order not to abide by the argument that professors are the only type of academics who accumulate entrepreneurial knowledge, engage in academic entrepreneurship, or move internationally (Siegel and Wright 2015). Pagani et al. (2019), for example, argue that students can have the same international experiences as faculty and may contribute to increases in knowledge stocks upon their return home. Our definition of the academic is thus broad.

Table 2 Research sample characteristics
Full size table

Data Collection and Measures

At the beginning of the survey, to reduce possible cognitive recollection problems and to assure accurateness and representativeness of the information provided, we asked respondents to take into account a five-year reference period. We provide an overview of our constructs of interest and how they were measured in the Appendix (electronic supplementary material).

Independent and Mediating Constructs

To capture temporary international mobility, we asked how frequently the respondents engaged in a range of international experiences. These experiences are: study, give lectures, official academic visit, summer school, internship, employment, contract research, consulting, research collaboration, attendance to conference or workshop, and laboratory or facility use. To take into account the full spectrum of temporary international mobility experiences, we did not opt for a single measure of international mobility as, for example, used in The GlobSci Survey (Scellato et al. 2015). Similar to Edler et al. (2011), McEvily and Zaheer (1999) and Fuentes-Fuentes et al. (2010), we captured the academic’s network strength by a frequency measure, i.e. how frequently respondents communicated with their different international contacts. For interpersonal network size, we asked how the academic’s amount of international contacts increased or decreased. In both cases (strength and size), contacts could be friends, colleagues, supervisors, and industrial, academic, or governmental.

Dependent Construct

Prior studies have measured knowledge gains through simple mobility rates, i.e. inflows and outflows of academics. We agree with the view of Cañibano et al. (2011) that this approximation is rather inadequate as this measure does not actually capture knowledge gained. We therefore capture entrepreneurial knowledge through a separate construct. We asked respondents to indicate how their international experiences helped/contributed to different entrepreneurial knowledge components, a set-up similar to that of other scholars (e.g. Fernández-Pérez et al. 2015; Levin and Cross 2004). These components are: knowledge on how to start/run a viable business, how to identify new business opportunities, certain technologies or products, production of products, offering services, customer needs, and funding or financials. They reflect the four entrepreneurial knowledge categories identified earlier in this paper. For analysis, we transformed the items questioning the different entrepreneurial knowledge components into a count variable, which allows us to assess the degree to which someone received entrepreneurial knowledge. Based on prior research (e.g. Mannucci and Yong 2018), we adhere to the argument that the use of a count variable is more precise than distinguishing between the importance of separate entrepreneurial knowledge components. This approach is also consistent with the theoretical build-up, i.e. looking at entrepreneurial knowledge as a compound construct of heterogeneous knowledge components and with the view that international experiences, bringing along knowledge components, can be considered a cumulative phenomenon (Cañibano et al. 2008). Methodologically, the use of a count variable, with equidistant data points, as a dependent construct in PLS-SEM is supported (Hair et al. 2017).

Control Constructs

Next to the measurement constructs, we captured controls. For each type of experience, we asked whether the purpose behind the experience was mainly educational, scientific and/or commercial. One might say, for example, that travels with commercial ends produce entrepreneurial knowledge by default. For demographic controls, age and gender were used as single-indicator controls. University support and encouragement towards entrepreneurial endeavors was captured through two Likert-scale items (one on support and one on encouragement). To control for the presence of prior entrepreneurial knowledge, we captured if the academic created a company or knows someone in his/her close environment who did (binary). We also captured to which country the academic travelled the most over the past five years. Since this question translates into a categorical variable with one category per country, we aggregated the responses to a macro level (North America, Europe, and other) and created single-indicator dummies. In subsequent analysis, we omitted the “other” category as reference category.

Data Analysis and Results

Model Preparation

We conduct an exploratory factor analysis (EFA) to purify the international mobility construct. This reduces the set of indicators used for this construct to a smaller amount which avoids data redundancy and simplifies the model (Janssens et al. 2008). The conditions to conduct the EFA are met as (1) the use of ordinal scales, instead of commonly used interval and ratio scales, is justified since they also generate reliable results, (2) the minimum amount of observations is met (ten times the amount of indicators), and (3) the indicators involved are sufficiently correlated to one anotherFootnote 5 (Janssens et al. 2008). We conduct the EFA using the principal axis factoring method, which is a popular estimation method providing a parsimonious representation of observed correlations between the indicators (de Winter and Dodou 2012). This method is suitable in our study as we did not have any predefined factor structure in mind and as it allows for subsequent SEM through the use of latent variables. The number of factors to retain was not stated a priori. Based on the combination of several criteria,Footnote 6 three factors are retained and deemed meaningful.Footnote 7 Typically, several activities and experiences are combined during an academic’s travel abroad (e.g. a research meeting amongst scholars can take place next to the participation in a conference or teaching at a summer school). Our EFA takes this into account by opting for an oblique rotation method (promax) and allowing for correlation amongst factors after rotation.Footnote 8 We analyze the resulting rotated component matrix and note that all factor loading values meet the required condition.Footnote 9 We extract factor scores (INT_FAC1, INT_FAC2, and INT_FAC3) and integrate them as indicators of our international temporary mobility construct in our PLS-SEM model for subsequent analysis. An overview of the descriptive statistics of our constructs can be found in the Appendix (electronic supplementary material). These statistics should be interpreted in light of conducting an SEM; not in light of traditional regression analysis. For example, PLS-SEM does not make assumptions about data distribution in its analysis (Hair et al. 2017). The use of non-normal data is even stated as one of the main reasons for its application. The use of the bias-corrected and accelerated bootstrapping routine in PLS-SEM waves this issue to some extent, as it adjusts confidence intervals for skewness (Efron 1987; Hair et al. 2019). However, in order to obtain reliable results, SEM requires other criteria to be fulfilled. We discuss these next, through the development of the measurement and structural model.

Measurement Model

We use a measurement model to evaluate the relationships between indicator and their corresponding construct (Hair et al. 2014). Our model uses reflective indicators which represent all possible items within the conceptual domain of the construct (Diamantopoulos and Winklhofer 2001). As a result, causality flows from the construct to the items (Wong 2013). Items can be used interchangeably and can be omitted without losing the meaning of the construct (Hair et al. 2014).

We run the PLS-SEM algorithm taking into account suggested convergence and iteration criteria.Footnote 10 A first step in analyzing a reflective measurement model is to evaluate the reliability and validity of the construct measures (Hair et al. 2014). The evaluation of reliability consists of checking indicator reliability and internal consistency reliability. Tables 3 and 4 show that both the outer loadings of the construct items and the composite reliability values meet the commonly used 0.70 or higher threshold (Wong 2013; Hair et al. 2014). Our construct items are sufficiently distinct and reliable.

Table 3 Indicators’ outer loadings and variance inflation factors
Full size table
Table 4 Constructs’ reliability, average variance extracted, and HTMT assessment
Full size table

We assess validity through convergent validity and discriminant validity (Hair et al. 2014). Support for convergent validity is provided when each construct’s average variance extracted (AVE) is 0.50 or higher, which would mean that the constructs explain more than half of the variance of the items (Bagozzi and Yi 1988). This is the case for our constructs (see Table 4). Discriminant validity represents the extent to which constructs are empirically different from each other. While previously assessed through the Fornell and Larcker (1981) criterion,Footnote 11 Henseler et al. (2015) show that this approach does not reliably detect the lack of discriminant validity in common research situations. They suggest the use of the heterotrait-monotrait ratio of correlations (HTMT), which is based on the multitrait-multimethod matrix. As our HTMT values are below 0.90, discriminant validity is established (Table 4).

Structural Model and Results

A structural model displays the relationships between the constructs being evaluated (Hair et al. 2014). Once reliability and validity of the measurement model is established, we take several steps to evaluate the hypothesized relationships between constructs. First, we test the structural model of potential collinearity and common method bias issues. The collinearity assessment is based on variance inflation factors (VIF) laying between 0.2 and 5. To assure no contamination by common method bias, VIF values should additionally be below 3.3 (Kock 2015). Table 3 shows that these conditions are satisfied; our structural model is free of collinearity and common method bias.

Instead of using traditional goodness-of-fit indexes, the assessment of the structural model is based on its ability to predict the endogenous constructs (Henseler and Sarstedt 2013). To this end, we analyze the path coefficients, which represent the hypothesized relationships linking the constructs (Hair et al. 2014). Estimates are shown in Table 5. These estimates are standardized on a range from −1 to +1, with coefficients closer to +1 representing strong positive relationships and coefficients closer to −1 indicating strong negative relationships. A bootstrapping procedureFootnote 12 allows to obtain standard errors to test for significance (Wong 2013). The results of this procedure are shown in Table 5 and Fig. 2.

Table 5 Path coefficients’ estimates and significance tests
Full size table
Fig. 2
figure 2

Visualization of p-values (values on arrows) and significant relationships (green arrows) in our model

Full size image

We find that the direct effect of international academic mobility on entrepreneurial knowledge is insignificant (p-value of 0.591). We reject Hypothesis 1: exposure alone to international experiences does not significantly influence the academic’s entrepreneurial knowledge. Looking at the indirect mediated effect through the academic’s interpersonal social network, we find that one’s network size fully mediates the relationship between international mobility and entrepreneurial knowledge. We thus accept Hypothesis 2b (p-value of 0.065) but reject H2a (p-value of 0.973). For completeness, coefficients of determination (R2) are shown in Fig. 3 as well. However, we should note that the R2 value of EK should be interpreted cautiously since count variables typically display low levels of explanatory power.

Fig. 3
figure 3

Disjoint two-stage approach model (p-values on arrows)

Full size image

Hierarchical Component Model

Arguably, our social network strength and size constructs can be operationalized on a high level of abstraction (Hair et al. 2014). This level of abstraction makes sense from a theoretical perspective since strength and size form relational and structural properties of one’s social network (Phelps et al. 2012). We create a single multidimensional higher-order construct on interpersonal social network (SN). This reduces our model’s complexity and averts confounding effects in multidimensional model structures, like multicollinearity. Since, theoretically, our SN construct is of formative nature, i.e. is formed by the strength and size properties, our hierarchical component model constitutes a reflective-formative Type II model (Becker et al. 2012).

Different methods are used to construct and evaluate a hierarchical component model (Sarstedt et al. 2019). We perform the disjoint two-stage approach (DTSA)Footnote 13 (Becker et al. 2012), since our social network construct (SN) also operates as an endogenous construct in our mediation model. Particular to this approach is that latent variable scores resulting from the prior analysis are saved for the STRENGTH and SIZE constructs and included in the subsequent hierarchical component model as indicators (LV_STRENGTH and LV_SIZE) of the aggregated SN construct. Estimation of the DTSA demands a Mode B – instead of a Mode A - indicator weighting modeFootnote 14 (Sarstedt et al. 2019). This choice relates to the formative character of the relationship between our higher-order construct (SN) and its two properties.

Assessment of the DTSA structural model happens on the grounds of the stage two results (Sarstedt et al. 2019). We find that the measurement model is not negatively affected by collinearity (VIF value of LV_STRENGTH and LV_SIZE is 1.380). Both indicators significantly affect the SN construct (p-values < 0.05), but the influence of LV_STRENGTH on the construct is larger than that of the LV_SIZE indicator. Estimates and results of path coefficients are shown in Table 6 and Fig. 3.

Table 6 Path coefficients’ estimates and significance tests of hierarchical component model
Full size table

In our aggregated hierarchical component model, we find that the mediation effect of one’s social network in the relationship between international mobility and entrepreneurial knowledge turns insignificant (p-value of 0.137). The same happens for the direct effect of social network on entrepreneurial knowledge (p-value of 0.130). This result is explained by the (slightly higher) importance of the strength property (compared to the size property) forming the social network aggregated construct. This shows that, although the strength property is relevant and meaningful in explaining an aggregated social network construct, it negatively influences the relationship between international mobility and entrepreneurial knowledge.

Control Variables

We test whether the addition of control variables has an effect on observed relationships and constitute alternative explanations to our model (Spector and Brannick 2011). First, our demographic controls, age and gender, did not affect our model significantly. Second, the model may be affected by the amount of entrepreneurial support or encouragement that the academic perceives by his/her home institution. Indeed, university support has a positive effect on entrepreneurial knowledge (p-value of 0.001). In addition, the indirect mediation effect through one’s social network size turns insignificant. Third, we control for the amount of prior entrepreneurial knowledge an academic might possess. While having started a company or knowing someone who did has a significant effect on entrepreneurial knowledge, the indirect mediation effect through one’s social network size remains significant at the 90% confidence interval (p-value of 0.082). Last, we check if one’s entrepreneurial knowledge is affected by the dominant travel destination of the respondent. We find a significant effect for academics who travelled to North America (p-value of 0.000), while the effect for academics who mainly travelled to Europe was insignificant (p-value of 0.111). In both cases, the indirect mediation effect through one’s social network size remains significant at the 90% confidence interval (p-value of 0.088). We summarize these findings in Table 7.

Table 7 Overview of significance tests on control variables
Full size table

Additional to these controls, we investigate the reasons for academics to be internationally mobile in the first place. One might argue that academics who travel abroad with commercial intentions in mind, are more prone to receive entrepreneurial knowledge than peers who do not show similar intentions. Our data shows that commercial aims were only reported by 3.7% of the academics, which is an average over all types of international experiences. This low percentage waves the argument that our model is biased by academics actively searching for entrepreneurial knowledge.

Discussion of Results

We find that the size of the academic’s interpersonal social network fully mediates the relationship between international mobility and entrepreneurial knowledge. This finding is in line with our premise that the academic’s social network plays a crucial role in the transfer of knowledge while being abroad. This finding relates to studies pointing to social capital and network formation as ingredient for the accumulation of internationally gained resources (e.g. Bauder 2020; Drori et al. 2009; Hoch 1987). With respect to the kind of entrepreneurial knowledge received, we observe that Chinese academics mostly report to have gained knowledge on how to identify opportunities (23.8%) and on how to start a viable business (21%), followed by knowledge on offering services (16.7%), on customer needs (15.7%), production (12.1%), and finances (8.2%). Chinese academics mostly gain market and organizational knowledge and to a lesser extent product and financial knowledge (Shane 2002; Widding 2005).

Mere exposure to international experiences, i.e. a direct relationship, does not influence the academic’s entrepreneurial knowledge. We find this slightly surprising, since one could, for example, argue that attending lectures or presentations at conferences directly provides knowledge relating to some of the entrepreneurial knowledge components and thus extend one’s entrepreneurial knowledge. However, our model does not take into account specific or different country and/or institutional aspects that can influence one’s entrepreneurial knowledge, even without frequent or in-depth interactions with an international network (Busenitz et al. 2000). In particular to the international context, an academic’s cultural background may limit or shape the amount of knowledge captured, how that knowledge is perceived (e.g. in terms of trustworthiness), and how important it becomes in one’s knowledge base (Dietz et al. 2010).

Our analysis also points to the particular importance of the structural size property over the relational strength property in the academic’s interpersonal social network. Thus, in order to receive entrepreneurial knowledge through international temporary mobility, the academic should focus on extending and broadening one’s international network instead of deepening relations within one’s existing network. This means that it is especially the new international connections that extend the academic’s knowledge base with heterogeneous entrepreneurial knowledge components (van Rijnsoever et al. 2008; Hoang and Antoncic 2003; Rasmussen et al. 2015) and not the existing connections providing other components through additional interactions. While prior research also notes that frequent interactions with informal contacts may provide heterogeneous knowledge, our study does not show similar result. However, we find this not very surprising due to the international context. Drori et al. (2009), for example, argue that while abroad people avoid the closer interpersonal contacts as these could constrain their access to novelty and heterogeneity in resources. The authors give the example of a Chinese transnational entrepreneur in North Ireland that saw no advantage to network with the Chinese community in Belfast. Although their study focuses on a priori entrepreneurs, the same reasoning applies in our study: academics, while abroad engage with new contacts to get the most out of their travels in terms of novelty and non-redundancy of knowledge received.

The mediation effect of social network size in the relation between international mobility and entrepreneurial knowledge only turns insignificant when university support is controlled for in our model. University support and encouragement towards entrepreneurship plays a crucial role in the enlargement of one’s entrepreneurial knowledge. Such support seems to outweigh the need for a broad international network. We believe that this is the result of entrepreneurship-related staff and courses being accessible and leveraged upon by academics that find themselves in so-called entrepreneurial universities, which translates into the availability of different entrepreneurial knowledge components (Wu 2010). Our results also point to the importance of the social network size property in receiving entrepreneurial knowledge even when academics already possess a certain amount of prior entrepreneurial knowledge, e.g. through having started a company or knowing someone who did. This result supports the creation and establishment of international networks outside one’s existing research group, university, or even region, i.e. so-called cosmopolitan networks (Bozeman and Corley 2004; Hayter 2016), as such networks may provide additional knowledge and ideas complementing individuals’ prior entrepreneurial knowledge bases, which typically hold localized knowledge.

Conclusion and Limitations

This paper studies the relationship between international temporary academic mobility and entrepreneurial knowledge. In particular, it investigates the role of interpersonal social networks, in terms of their size and strength property, in this relationship. We do not find a direct relationship between international mobility and entrepreneurial knowledge. Without interactions abroad, i.e. when isolating oneself, academics will not enlarge their entrepreneurial knowledge base. However, in terms of the indirect relationship through one’s social network, we find that the size of the academic’s social network fully mediates the relationship between international mobility and entrepreneurial knowledge. New international contacts extend the academic’s knowledge base with heterogeneous entrepreneurial knowledge components. Our research provides support to the idea that academics reap benefits from their internationalization efforts, adhering to the brain gain discourse. Previous studies have reported positive effects of international mobility on (scientific) productivity, career development, and knowledge transfer activities (e.g. Veugelers and Van Bouwel 2015; Edler et al. 2011). Our research now complements these studies with the insight that academics, while being abroad, increase their entrepreneurial knowledge through the expansion of their interpersonal network.

Our study has several limitations which constitute interesting pathways for further research. The first limitation concerns our data and methodology. While PLS-SEM is suitable to study cause-effect relationships through a mediating construct (Hair et al. 2017), we do not argue causality in our study. To build a convincing case for a causal relationship, we should demonstrate a temporal sequence by which the presumed cause precedes the presumed effect. Our data is limited towards such demonstration. Our data also builds upon a sample that represents only 1.2% of the studied population. Although we compare our sample to related research, this limits the representativeness and generalizability of our study. As path analysis and SEM are getting more popular, the use of controls in structural models, and especially in mediated models, needs further attention (Atinc et al. 2012). Second, we operationalized our model with a focus on the importance of heterogeneity in the entrepreneurial knowledge base. While this approach is supported by existing literature, some scholars (e.g. Miralles et al. 2016) point to the particular importance of combining heterogeneous and homogenous knowledge components to form one’s entrepreneurial knowledge. While we support this argument, we argue that coordination and recombination of homogenous and heterogeneous knowledge parts is crucial to the identification of entrepreneurial opportunities, which constitutes a first step of the entrepreneurial process (D’Este et al. 2012). Our study is limited to study the perceived receipt of entrepreneurial knowledge (Levin and Cross 2004) and does not make statements on the potential subsequent engagement of the academic in the entrepreneurial process. In that regard, we also limit ourselves to knowledge as a key resource and discard other individual level resources and capabilities (e.g. networking capability). Third, since Chinese academics tend to travel mostly to the West for international temporary mobility experiences - which shows in our data - our findings may be limited to a setting in which academics from an emerging country travel to developed countries (Wright et al. 2005). The inverse setting may constitute an interesting path for future research. We also question why academics travel to certain destinations in the first place and if/how this choice relates to the (entrepreneurial) knowledge acquired. Last, our study does not take into account any domestic academic mobility or the complementary or substitute role domestic networks could play in our model. For example, due to proximity aspects, we would expect the role of frequent interactions to be more prevalent in a domestic academic mobility setting. Also, the existing domestic interpersonal networks may play a crucial role in the return of the academic to the home country (Bauder 2020; Baruffaldi and Landoni 2012).