Individual versus Institutional Collaboration

Research collaboration can be defined as “social processes whereby human beings pool their human capital for the objective of producing knowledge” (Bozeman, Fay, and Slade 2013). A recurring theme in work on research collaboration is the relative importance of organizational versus individually-driven collaborations, with studies generally concluding that both are important.

One body of literature trying to make sense of this duality relates the distinction between organizational versus individual drivers to the nature of the science conducted. Thus Chompalov and Shrum (1999) find that certain characteristics of research projects (namely project formation, magnitude, bureaucratic organization, interdependence, communication, participation, and technological practice) not only span scientific fields, but also result in a useful classification of how collaborative projects are managed. In particular, what they term “technological practice,” the development of instrumentation specifically for the project, is useful in differentiating among four different types of approaches with different levels of institutional control. But it is noteworthy that of the four types, the formal institutional arrangement plays an important part only in the “managerial” approach. In a subsequent study, Chompalov, Genuth, and Shrum (2002) find that greater interdependence in data acquisition, analysis, and communication of results is correlated with less formal structures and greater success. They point out that if data are jointly obtained and verified by collaborators, a hierarchical arrangement becomes less important. Similarly, in a study of six fields of science, Wagner and Leydesdorff (2005) report that as ICTs and communication gain importance in collaboration, the selection of partners and the location of research is decided by the individuals rather than institutions or organizations.

Corley, Boardman, and Bozeman (2006) make a similar point using a different paradigm. They argue that collaborative research projects take place within the context of both epistemic norms and a certain type and level of development of organizational structure. In the case of epistemic norms, the definition of a scientific field could be seen as an indication of a high level of maturity; its organizational counterpart would be the establishment of an academic department. Collaborative research projects differ in terms of the level of development of the epistemic norms and/or organizational structure within which they operate, and Corley et al. (2006) argue that the two interact in a predictable way. When the epistemic framework is underdeveloped, there is a need for a stronger organizational structure, but when the epistemic dimension is already well-developed, a shared, common “language” reduces the need for strong organizational management.

As suggested by Nelson (2005), knowledge of such a language is an important “social technology.” Scholars from less developed countries are likely to find it relatively easier to participate in projects in already defined fields where a shared scientific language is well-established and can act as a coordinating mechanism. This is particularly likely for scholars who were trained abroad and know those norms, but even when a scholar from an underdeveloped context is not entirely au fait with the norms of the field, s/he is likely to learn them more quickly to the extent that they are more strongly shared.

But where there are not strong epistemic norms, Nelson argues that coordination and governance may present a special challenge for scholars from institutionally underdeveloped contexts. He points out the complex array of tasks that support academic endeavor – establishing an appropriate division of labor; management, control, and coordination; hiring, rewarding, and occasionally releasing labor; financing and so forth. Developing those is both a prerequisite for and a desired outcome of effective indigenous research.

Relatively little work has been done on how individual versus institutional considerations vary for the establishment of research collaborations from more versus less developed research contexts. However, evidence from a related field, university–industry linkages, suggests that in contexts with weaker institutions, the role of individual drivers becomes more important. Thus a study of Brazil has shown that personal relationships are an important mechanism for university–industry linkages in emerging industries where public knowledge and support are limited, and less so in mature industries (Bodas Freitas, Marques, and Silva 2013). A similar pattern of relying on personal ties when institutions are weaker has been found in Italy where small firms rely on personal contacts to establish university–industry linkages, whereas large, vertically integrated firms use institutional arrangements (Bodas Freitas, Geuna, and Rossi 2013).

This suggests that individuals and individual relationships play an important role in establishing research collaborations in less developed countries. In the case of research collaborations with international partners, it is likely that useful practices from more developed contexts will be borrowed. In the process, scholars from a less developed context stand a chance of contributing not only to the local knowledge base, but also to its institutional development.

Learning from Local and Global Connections

Traditionally, global connections occurred among developed countries such as Japan, Europe, and the United States with a technology-seeking aim (Song and Shin 2008); recently, however, there has been an increase not only in the inflows of knowledge and FDI to developing economies (UNCTAD 2013) but also the emergence of outflows from those countries (Cuervo-Cazurra 2007; Gammeloft, Barnard and Madhok 2010; Álvarez and Marín 2013). Global connections in the literature on less developed countries can be categorized as either “inbound,” where not only the desired new knowledge, but also much of the initiative for obtaining it comes from abroad, or “outbound,” where the desired new knowledge lies elsewhere, but the initiative for seeking it is local. The case when advanced MNCs set up and nurture subsidiaries in less developed contexts (Narula and Dunning 2000, 2010) is perhaps the best documented example of inbound global connections. And when emerging MNCs (EMNCs) engage in FDI in especially more developed contexts (e.g., Nam and Li 2013), those global connections can be described as outbound.

There are benefits but also problems associated with both those paths. There is a growing body of evidence that outbound connections may not yield the desired benefits. The attempts of EMNCs to source knowledge and capabilities through establishing global connections do not always succeed. In particular, they seem to some extent excluded from the more tacit sources of knowledge, and better able to source market-based assets (such as more codified knowledge) (Barnard 2010) or operate in an incremental, exploitative way (Rabbiosi, Elia, and Bertoni 2012). Furthermore, EMNCs face the challenge of overcoming the liability of foreignness, that is, the cost of doing business abroad (Barnard 2010), and overcoming the disadvantage of weaker capabilities located behind the technological frontier (Ramamurti 2009; Álvarez and Marín 2013); for example, the average asset base of EMNCs is only 15% that of the MNCs of the developed world (UNCTAD 2013).

As for advanced MNCs, there is also considerable evidence that the host context does not always benefit from the entry of the MNC (Meyer and Sinani 2009). The reasons given that global connections fail to facilitate technological advance are often that the local context lacks the capability to take advantage of the knowledge, for example, because the social ties that would enable knowledge sharing to take place are limited (Eapen 2012), or because the host country lacks the sufficient absorption and adaptation capabilities (Kumaraswamy et al. 2012). Another common explanation is that the MNC does not extensively share knowledge with the subsidiary in the developing country, generally because the mandate of the subsidiary is limited, for example, when the MNC sought out the location for its low wages, and consequently located mainly low value-adding activities there (Sethi et al. 2003), but sometimes also because of political considerations inside the MNC (Becker-Ritterspach and Dörrenbächer 2009).

We argue that an overly simplified view where parties are seen as either “providers” or “receivers” of knowledge is problematic. Moreover, in such a view, less developed countries are typically seen as receivers rather than creators of knowledge. With the stark conceptual divide between receivers and providers of knowledge, it is hard to see the possibilities of feedback loops and mutual benefit. This analytic oversimplification makes it hard to understand the role of both local engagement and global connections in upgrading.

It is telling that this analytic oversimplification is less often found in studies of the global connections of advanced economies. In those studies, there is generally an appreciation for the fact that different countries have different specializations, and that entities in both contexts stand to benefit from sharing knowledge (Cantwell and Piscitello 2002, 2005). Studies have explicitly examined the benefits not only to the “recipient” (e.g., Chung and Alcácer 2002), but also to the “provider” (Singh 2007; Yang, Phelps, and Steensma 2010) of knowledge, and consistently find that knowledge does not flow only one way. Bartlett and Ghoshal (1998) conclude that the interactions between MNC and host countries are much more complex and interactive than simply establishing “replicas” abroad and in fact are close to the concept of an international network. Seen in this way, a more internationalized value chain (Kaplinsky 2004) ensures that host countries exploit competencies over the firm’s network and also develop new competencies and knowledge resources (Rugman and Verbeke 2001; Yang, Mudambi, and Meyer 2008). Singh (2007) finds not only that the knowledge outflows from the host country back to the MNC in more developed countries (or more developed sectors) are greater than their knowledge inflows to the host country, but that even in the case of relatively less developed countries or sectors, outflows to the MNC almost equal the knowledge inflows from the MNC.

Although it is generally true that the capabilities of less developed countries are more limited and their institutions less developed than those of the leading economies, Marin and Bell (2006) provide evidence that the most beneficial global connections are those where the “inbound” and “outbound” distinction is less clear. In their study, they found that innovation in Argentinean subsidiaries of advanced MNCs was especially likely when the subsidiary was connected to the headquarters of the MNC but also to local providers of knowledge and expertise.

In contrast to the elusiveness of technological upgrading when the source of knowledge is distinct from the recipient, the evidence from Marin and Bell (2006) suggests that the best way for learning to take place is when both parties play an active role as partners and collaborators in the process of learning and knowledge production. For example, Rasiah (2003) reports how Malaysia and Thailand, key exporters in the electronics sector, have improved their technology capabilities as FDI moved from just the expansion of production to the development of technology. Indeed, evidence about the ongoing importance of absorptive capacity (Goode 1959; Cohen and Levinthal 1990) in its original conceptualization (i.e., with the emphasis on the considerable effort needed to be able to benefit from knowledge spillovers) suggests that the notion of a distinct “sender” versus “receiver” of knowledge is problematic. The process seems more robust when both parties are recognized as potential sources of knowledge. Thus, to advance research in a developing country, it is not enough for the role-players from the developed world to be highly competent; their counterparts from the developing world itself should be too (Barnard 2010).

The arguments put forward in the literature on knowledge flows within and across organizations apply as well to individuals. In fact most arguments, particularly those related to absorptive capacity, are rooted in theories of individual behavior, cognition, and psychology. Collaboration among scientists takes many forms, ranging from informal advice to joint research and co-authorship. Most scientific collaboration studies investigate co-authorship relations. A consistent finding is that scientists with higher reputation, essentially having more publications, attract more collaborators (Newman 2001). In other words, there is a strong tendency of young scholars to connect with scholars of high reputation and, at the same time, a strong tendency among highly visible scientists to collaborate (Verspagen and Werker 2004). Thus, many co-author relations are marked by a strong differential in terms of experience, and hence knowledge might flow predominantly in one direction. However, since by definition every author contributes to the joint scientific paper, knowledge flow will never be strictly one-directional; and even less so when we differentiate between generic knowledge (mainly from supervisor to student) and specific knowledge (mainly from student to supervisor). Knowledge differentials, averaged over specialties, between “world leading” scholars are going to be small in general, which strongly suggests bidirectional knowledge flows. When there is a “world leading” collaboration between scholars across more and less developed contexts, we may therefore expect both benefits in terms of the global production of knowledge, and also the development of indigenous knowledge and capacity.

This perspective suggests that scientific collaboration among scientists from developed and developing countries is essentially a win-win situation. How could the developing country then be negatively affected by numerous collaborations spanning the global and local science system? In the debate among scholars and policymakers two concerns are frequently raised.

The first concern is that excessive international collaborations might result in a local science system that fits the science systems of developed countries well but is maladapted to the local context. Developing countries may have research agendas that do not coincide with those of the leading developed communities because they are endowed by different resources and have different societal needs. This confrontation of research priorities has been extensively reported in the case of Asian business schools, where the research agendas, driven by developed country contexts, struggle to adapt to local circumstances, and thus fail to tackle issues that concern local managers (Meyer 2006; Leung 2007). This, however, is not unique to emerging countries, and has been also found in advanced economies such as the Netherlands, where the fruits of research bring prestige but do not bring profits to the region (Rip 2002). Based on similar arguments, Pouris and Ho (2014) express concern that the extensive collaborations of African scholars may occur at the cost of the development of local and regional research systems.

The second issue originates from the fact that developing countries are marked by strong heterogeneity. The local science system includes few scientists and institutions of research excellence. Most scientific institutions lag largely behind the scientific frontier as they lack the necessary resources. Only the few centers of excellence in the local science system are able to engage beneficially in international collaborations at the scientific frontier. Thus, supporting global connections means that one needs to invest more in centers of excellence rather than in the broader array of institutions. Individual scientists of excellence might then devote most of their collaboration time and resources to international collaborations and collaborate less with their less privileged local colleagues.

The concern then is that the centers of excellence integrate within the global science system but at the same time disconnect from the rest of the local science system. Instead of upgrading the overall indigenous science system, fostering international collaborations would increase heterogeneity, and weaken coherence, within the indigenous science system. Connectedness with the global community of science may help advance science, but only in small enclaves, while developing indigenous public research capability may contribute to a broad base of researchers, but operating largely behind the technological frontier.

The literature documents several reasons why globally connected but locally disconnected enclaves may develop around the most competent researchers in a developing country (see, e.g., Barnard, Cowan, and Müller 2012). Such enclaves have been previously documented in international business contexts in less developed countries, especially in those cases when quality is of particular importance (Akbar and McBride 2004; Feinberg and Majumdar 2001). In an earlier study of scientific collaboration, we found that the world-leading scientists in a developing country – rather than scholars whose emphasis is essentially local – tend to be well connected internationally, but at the same time remain strongly connected locally (Barnard et al. 2012). Thus, the concern of enclave formation could not be supported empirically in that study. On the contrary, “world-leading” scholars with a rich international collaboration network appeared to be important in the diffusion of global knowledge locally.

To this point our discussion has followed a stylized image where knowledge flows through scientific collaborations among scientists that are either from the developed or from the developing world. This picture is refined in the next section by taking into account that scientists are actually highly mobile between the local and global science system.

How Mobile Scientists Connect Global and Indigenous Science Systems

The analysis of the process of knowledge creation and diffusion cannot be separated from the study of individual scientists. On the one hand, Fischer (2001) points out that it is individuals, not organizations, who are the generators of knowledge; on the other hand, a large part of this knowledge is embodied in individuals as tacit knowledge (Polanyi 1983). The localization of knowledge flows is thus tied to and can be explained through local mobility of knowledge workers or inventors (Breschi and Lissoni 2009; Møen 2005).

Extensive literature suggests that knowledge diffusion occurs to a large extent through the movements of individuals (Saxenian 1996; Argote and Ingram 2000; Rosenkopf and Almeida 2003). In the context of inter-firm knowledge transfer, Malecki (1997) argues that individuals are an important channel of transfer. In the semiconductor industry, for example, studies have found that when inventors move from one firm to another, they carry knowledge from their old firm to the new one, affecting the subsequent patenting activity (Saxenian 1990; Almeida and Kogut 1990; Song, Almeida, and Wu 2001). Furthermore, organizational learning through employee turnover, or learning-by-hiring, is a significant driver of mobility in innovative firms (Cassiman and Veugelers 2006) and high-tech industries (Song, Almeida, and Wu 2003; Palomeras and Melero 2010); for example, the hiring of star scientists (Schiller and Diez 2009), and in the case of pharmaceutical industry, the adopting of biotechnology (Darby and Zucker 2001).

Similarly, in the context of science, PhD training, exchange of researchers, and professorship appointments are important ways of transmitting not only scientific knowledge but also social knowledge, as discussed for example by Nelson (2005). This has been pointed out at least since Polanyi (1958). More recent is the discussion of Scherer (2000) on how the emigration of many Austrian and German economists to the United States during the Third Reich affected American economic development. Similarly, Kogut and Macpherson (2011) provide evidence that during the 1990s economic paradigms (e.g., on privatization) traveled from the United States to Europe through mobile scientists.

Thus knowledge moves together with the individuals who hold that knowledge. This idea is at the origin of the brain drain discussion, which started in the 1960s when scientists and other knowledge workers trained in the United Kingdom migrated to the United States (Cervantes and Guellec 2002). The same label has also been used to describe the phenomenon in which untrained knowledge workers migrated from developing to developed countries, mainly the United States, in order to receive training. In the former case the loss is in actual human capital (the most skilled leave), while in the second it is the potential human capital (the brightest leave) (see, e.g., Gaillard and Gaillard 2001). Meyer (2001) reports that as the academic system and research network connections are more complete, the first kind of migration, the loss in actual human capital, is more common; for instance, having a more complete techno-economic network, South African expatriates tend to go abroad when they are more qualified, while Colombian expatriates, with short and limited networks at home, emigrate with only a bachelor degree, seeking to pursue further studies abroad.

There have been many policy initiatives of developing countries since the 1970s that aimed to repatriate skilled expatriates, through for example compensation or taxation schemes (Meyer 2001). Most of these policies failed, except perhaps for some cases in Far East (e.g., South Korea, Singapore, and Taiwan) (Gaillard and Gaillard 2001). This led policymakers in the 1990s to acknowledge (or hope?) that even when expatriates stay abroad they may bring some value for their country of origin, so that they should not be considered a loss but rather an asset that could be mobilized (Meyer 2001). As a consequence, network initiatives emerged to create networks among and with expatriates in order to create social capital (Gaillard and Gaillard 2001; Meyer 2001). Brown (2000) reported the creation of more than 40 networks in developing countries with the purpose of bringing closer expatriates’ skills to their country of origin (e.g., SANSA initiative of South Africa or CALDAS in Colombia). In recent years, however, studies have reported a decrease in the level of permanent migration (Koser and Salt 1997; Cervantes and Guellec 2002), temporary mobility being far more common; especially among PhD students and post-docs with the aim of gaining knowledge, experience, and contacts that can be exploited upon their return to the home country (Edler, Fier and Grimpe 2011).

These migration patterns and the choice of location are determined by the social (Li et al. 1996), professional (Mitchell 2000; Beaverstock 2005), and co-ethnic (Meyer and Brown 1999; Saxenian 2002) relations that an individual has. Melin (2004) reports that when post-docs move they develop a richer network but often they move along established connections of their senior scientists at their home institution. Therefore, how scientists move is both a result and a driver of the collaboration network (Scellato, Franzoni, and Stephan 2012; Cowan, Feldman, and Kogler 2006).

Overall, the reason behind the temporary mobility of scientists is the perception that being part of the global network and collaborating with scientists abroad improves capabilities and research productivity (Edler 2007; Defazio, Lockett, and Wright 2009). Jonkers and Tijssen (2008), in an empirical study of China, and Scellato et al. (2012), in a study of 16 countries, show that mobility fosters collaboration networks and that scientists who experienced foreign stays have richer international collaboration networks and higher scientific standing. Both studies report a positive correlation between foreign experience and the number of international co-publications. Moreover, in a study of Swedish scientific mobility, Melin (2004) finds that a post-doc abroad has important integrating effects between the home and the host country.

Based on these ideas, we argue that scholars in developing countries who have had extensive experience working and training with scholars from the leading knowledge producers will play a special role in development of indigenous research. Ynalvez and Shrum (2011) point out in a Philippines-based study that benefits could occur because of training which would ensure that they are familiar with the norms for quality research in their field, but also because of the value of foreign socialization and developing contacts. However, developing relationships with scholars abroad can occur in many different ways. Research partners may be diasporans who are keen to establish collaborations with scholars at home, collaborators may have met at conferences, or a third party with extensive international experience may broker the relationship. Either way, research collaborations between local researchers from the developing country context and foreign scholars are likely to play an important role in the development of local science. In a study of the returning process to China of scholars and students overseas in the period 1997–2001, Zweig, Changgui, and Rosen (2004) found that returnees holding foreign, and to a lesser extent domestic, PhDs helped domestic PhDs to establish global networks; they also increased their human capital and attracted more grants. This phenomena of collaboration has also been observed in developed nations like Canada, where academic Canadians living abroad have been found to have an active role in Canadian knowledge creation, and thus, not constituting a pure brain drain (Cowan et al. 2006).

The extant literature on the mobility of scientists thus emphasizes the effect of mobility on the number and quality of scientists in a system (the human capital aspect) and how mobility relates to social connections within and among systems (the social capital aspect). Both international mobility and international collaborations foster global connectedness and affect the development of indigenous science in middle-income countries. To illustrate this hypothesis in one context, the next section presents an illustration of the importance of global connectedness in South Africa. South Africa is a useful setting and a representative example, because its level and patterns of scientific expertise are typical of middle-income countries generally. Thus, the results obtained in the empirical study, even though restricted to the case of South Africa, represent in general the issue of connectedness in developing countries and are intended to illustrate the theories presented in the first part of this chapter.

Relative Positioning of the South African Science System

A relative ISI citation ranking from 2013 reveals that most of South African science is behind the technological frontier. Table 18.1 indicates that the impact of most South African research is similar to the impact of work done in peer countries, other leading emerging economies like Argentina, Brazil, Saudi Arabia, and Mexico. Even in the best performing disciplines, the work of most South African researchers remains quite far from the frontier, and has a comparable impact to that of researchers in the non-core disciplines of small late-coming economies like Finland, Singapore, and Taiwan.

Citation rates are a commonly used measure of the impact of research and used to establish quality in a variety of knowledge production contexts (DuBois and Reeb 2000; Jaffe 1989; Starbuck 2005). Although there is evidence that citations reflect a social as much as a cognitive process (Leydesdorff and Amsterdamska 1990), it is a useful high-level indicator of the relative prominence, impact, and arguably also quality of research.

Because citation rates vary by discipline, we show the South African citation rate for each discipline separately (see Table 18.2.) There is evidence of a number of centers of excellence, such as Clinical Medicine and Immunology, related to the HIV epidemic in South Africa, Space Science, which reflects the extensive government support for South Africa’s ultimately successful bid to host the Square Kilometer Array,³ and the small but highly cited body of work on Computer Science. However, the overall average citation rate of publications by South African authors in ISI is almost exactly the average citation rate of ISI publications overall.

It is likely that other middle-income countries will exhibit a similar pattern: a few centers of excellence that relate mainly to local conditions, but generally scholarly work has “average” impact with a small community of researchers. Clearly, such countries will stand to benefit from being connected to scholars from the global technology leading countries.

Data Source

We employ a unique data set made available by the National Research Foundataion (NRF) of South Africa. The NRF is one of the main government research funding agencies, and has as its mission to promote research and develop research capacity. As part of the process by which it awards research grants, the NRF has instituted a “rating” process: before applying for funding, any researcher must apply to be rated. The process involves submission of evidence of all past research output in addition to the candidate’s work history and other relevant details. The NRF assessment is based on roughly six referee reports, written by local and international scholars. The referee reports are used by an independent assessor and a specialist review panel to arrive at a rating for the applicant. A given rating is valid for a fixed period,⁴ and when it lapses the researcher must reapply. Reapplication may result in a change in rating, either up or down, so ratings are a good reflection of very recent performance.

Table 18.3 defines the different possible rating categories according to experience and quality of work. Conveniently, NRF categories explicitly observe where researchers’ work fits in the overall global or local context. A-rated and to a lesser extent B-rated researchers are seen as contributing to the international knowledge frontier, whereas the focus of C-rated researchers is primarily local. As the rating of scientists as world-leading scholars (A-rated), internationally recognized scholars (B-rated), or scholars acting within the local scientific community (C-rated) has strong implications for research funding, scholars, and their universities, are certainly motivated to achieve the highest possible rating. The data are therefore well suited to explore the local/global tension and potential in a technologically less developed context.

Sample

The study focuses on researchers working in science, engineering, and social science at South African universities. All NRF filings from the beginning of 2002 until the end of 2012 are used for the analysis. The data include socio-demographic information (age, sex, race, and marital status) as well as professional information (title, scientific field, and list of publications) on the researcher and the rating by the NRF. Our focal population for the analysis consists of the 2778 established scientists who received at least one valid rating (A, B, C, P, Y, or “Rating Unsuccessful”) during the sample period.

We construct a dynamic co-authorship network using peer-reviewed publications, evolving from 2002 to 2012 using a lagged five-year moving window, and use this network to measure the network position of focal researchers at the time of rating. Because the rating might occur any time in a given year, we consider the co-authorship network five years prior to rating. If for example a focal rating was obtained in 2002, we measure the network position on the network that accumulated from 1997 to 2001.

For the analysis it is crucial to clearly identify foreign co-authors and rated co-authors. Yet, some co-authors in the publication list might be South African but not rated. We identify non-rated but South African researchers based on further data. The quality of the procedure for the identification of foreign researchers has been checked and found satisfactory with errors in the lower single-digit range, and conservative with respect to rated and foreign researcher categorization. After the whole cleaning procedure, the complete network accumulated from 2002 to 2012 includes 36,633 researchers (32,361 foreign and 4272 rated South African) connected through 211,207 single or multiple co-authorships.

Network positions, such as the gatekeeping score, are measured on the weighted network. We follow the approach of Newman (2001) and weight each link that originates from a paper with n co-authors by 1/(n – 1). For example, a paper that is co-authored by three researchers adds three links of strength 1/2 to the network. The idea behind that weighting scheme is that each researcher invests a fixed amount of interaction time in each publication which is divided equally among all co-authors.

The Gatekeeping Function of Internationally Renowned Scientists

Our network measure of gatekeeping (Barnard et al. 2012) quantifies the extent to which an individual scientist occupies a position in the co-author network that is characterized by both strong research connections with international, non-South African scientists for knowledge acquisition and, at the same time, high centrality within the South African science system for knowledge diffusion.

The discussion on local and global connectedness in the literature argues that scientists may follow one of two strategies which tend to exclude each other. One strategy is to aim at international recognition by tackling research issues that are acknowledged within the international research community through extensive collaboration with international scientists. The other strategy is national. Here the researcher tackles research issues with more local relevance in collaboration with local scientists. Researchers that follow exclusively the international strategy would pursue international collaborations for the sake of international connectedness, while researchers that follow an exclusively “national” strategy tend to lose sight of international contacts. The result in both cases would be a poor performance in terms of our measure of gatekeeping. To capture these two key aspects – connection to the international knowledge frontier and connection within the South African community – we define the gatekeeping measure as follows:

The external weight of a researcher reflects his/her international connectivity and is captured by the (weighted) number of papers with international scholars. The knowledge diffusion capacity within the South African science system is measured through the average weighted reach of the focal researcher to all other researchers in South Africa. Average weighted reach is defined as the average of the inverse of the shortest weighted path from the focal agent to all other agents in the network. The higher the average weighted reach of the focal agent, the closer the focal agent is on average to other agents in the network. This is beneficial for knowledge diffusion because knowledge sent by a researcher spreads more easily and faster to recipients who are close in the network. Thus gatekeeping of a researcher increases as her external weight increases and as average weighted reach to other South African scientists increases.

Table 18.4 shows how the researchers’ gatekeeping score relates to their ratings. The second column, “Observations,” provides the number of observations in each category. A-ratings (world-leader) are relatively rare and C-ratings (local participant) are by far the most common. The roughly 20% of unsuccessful ratings includes cases where the track record of the applicant was either insufficient for rating or where a criteria for rating was not met (such as incomplete submission or main workplace outside South Africa).

The average gatekeeping scores, third column of Table 18.4, are decreasing from A- to B- to C-rating, and even lower for unsuccessful applicants. Since gatekeeping is composed of two factors, embeddedness in the local science system and access to the international science system, it is interesting to investigate the source of the gatekeeping advantage of higher rated researchers. Local embeddedness, that is, average weighted reach, is very similar across rating categories. Clearly, systematic differences in gatekeeping scores arise from differences in the strength of external co-authorship ties (see fourth and fifth columns “Average weighted reach” and “External weight”).

A positive correlation of higher ratings on the one hand, and higher gatekeeping scores (higher external weight) on the other hand is also observed when we focus on individual time periods or individual scientific disciplines.⁵ For example, the same pattern is observed for the sub-population of biologists. Biology is one of the fields where South African scientists are, according to ISI, recognized as just above the mean. In this sense the biology discipline is representative for the whole, and we are able to enrich the global statistics by an image of the network among biologists.

Figure 18.1 presents the main connected component of the co-authorship network among focal (i.e., senior rated) biologists, taking into account all peer-reviewed scientific articles between 2006 and 2010. The size of the nodes increases with the external weight (in logs). Visual inspection of Figure 18.1 suggests that global patterns described above are also reflected in this relatively small sub-sample of the network: A-rated researchers (black in Figure 18.1) are relatively well embedded (central in Figure 18.1) in the South African co-author network, and have more external collaborations (large node size in Figure 18.1); thus, taking important gatekeeping positions in the network.

The example puts some light on two more general issues discussed above. Firstly, we observe that higher ranked researchers have more international collaborations. This suggests that knowledge flows not only one-way, from the global to the local science system, but both ways. In particular, A-rated South African scientists contribute as world-leaders (by definition) to global knowledge in international collaborations. The second observation is that researchers with many international connections tend to remain connected within the local science system. Thus, a strategy of international connectedness does not necessarily imply a cutting-off from the local science system but may well complement local upgrading.