Introduction

What determines the level of connectivity of a peripheral economy to global networks of inventors? Global innovation systems are becoming increasingly complex, involving a wider range of locations^[1]. As value chains are disaggregated across borders, countries are increasingly interconnected in global innovation networks, part of which is reflected by the growing dispersion of patenting activities (Balconi et al., 2004; Breschi and Lissoni, 2009). Locations outside core OECD countries, attempting to catch up technologically with core developed economies, try to attract multinational enterprises to perform innovative activities in their territories and create linkages to these global innovation networks. Connectivity provides an economy with access to a wider variety of world-class pools of knowledge (Medina et al., 2020; Qiu et al., 2017).

The factors affecting the connectivity of core and peripheral regions have attracted some attention in the last few years (Cano-Kollmann et al., 2016; 2018; Glückler, 2014). However, there is still ample room to explore how peripheral economies create or source different types of knowledge, and what role organization-based linkages (i. e., “pipelines”) play in this process (Lorenzen and Mudambi, 2013). This is the main motivation for our paper.

We make two related arguments. First, it has been noted that value chain activities involving a high level of tacit knowledge “tend to remain more agglomerated in the parent company” (Cantwell and Santangelo, 1999: 101). These activities will be performed mostly by collocated teams or teams with members in global centers of excellence (Gittelman, 2007). Therefore, we argue that the local economy obtains limited connectivity to global innovation networks. Second, we argue that some leading innovative firms may possess complex organizational capabilities that give them the ability to orchestrate these such activities when they are dispersed over geographic space (Cantwell, 1995; Cantwell and Mudambi, 2011; Tallman and Chacar, 2011).

The connectivity of a location is defined as the particular configuration of its global linkages combined with the specific network structure of these linkages (Lorenzen and Mudambi, 2013). In general, linkages between locations can arise either from organizations or from individuals. Since we use patent data, we will mostly refer to organizations as “assignees” (the owners of the patent) and to individuals as “inventors” (the individuals listed as inventors in the patent). In the literature, organization-based linkages have been referred to as “pipelines” (Bathelt et al., 2004), while those arising from individuals have been referred to as “personal relationships” (Lorenzen and Mudambi, 2013). Further, locations differ in terms of the extent to which their linkages are concentrated in a few central actors. In this paper, we do not analyze personal relationships (i. e., networks between individual inventors working together without being affiliated to an assignee). We focus on organization-based networks, i. e., networks coordinated by assignees for which inventors work. We examine empirically one aspect of connectivity in detail: the geographical dispersion of organization-based pipelines across national borders, as reflected by the patenting activity.

Non-core locations have the most to gain from connectivity to global innovation systems (Abramovitz, 1986). In particular, these gains can be best leveraged by economies that have achieved some degree of maturity in terms of local innovative capabilities. “Peripheral economies” form a particularly important class of non-core economies. In the international business literature, the concept of a “peripheral” economy fills an intermediate category (Molero, 1995) in the rigid “developed vs. developing/emerging” economies dichotomy. Benito and Narula (2008) provide a definition that characterizes peripheral economies according to detailed criteria like levels of foreign direct investment (FDI), trade in intermediate and manufactured goods, and levels of innovation, in order to distinguish them from core OECD economies. Some southern and eastern European countries are good examples (Benito and Narula, 2008; Liagouras, 2010; Narula and Guimón, 2010). As value chains are disaggregated across multiple countries, this provides opportunities for non-core locations to participate in the high knowledge components of those global value chains (McWilliam et al., 2020). Further, since peripheral economies are likely to lag the core in terms of innovation capabilities in almost all sectors, connectivity is likely to have particularly strong effects for them.

We study one particular aspect of the global connectivity of peripheral economies, namely the international dispersion of inventor networks. These inventors are typically employed by organizations so that we capture organizational networks that are defined by relationships between these individuals that are part of inventor networks. We use Portugal and Greece as examples of peripheral economies (Benito and Narula, 2008; Narula and Guimón, 2010). Both countries can be considered peripheral to the core region of Europe and are comparable in size, income and the level of development of their innovation systems. Furthermore, their location in the perimeter of the European continent (Portugal in the southwest and Greece in the southeast), and the fact that they do not share borders with the core European economies, create similar challenges in terms of integration with the rest of the continent. We analyze patent data for both countries, encompassing all the patenting activity linked to Portugal and Greece. We include patents from local firms with local inventors, patents from foreign-based assignees with local Portuguese or Greek inventors, and patents from local Portuguese or Greek firms with foreign-based inventors. Therefore, our sample includes the patenting networks of firms and inventors located in 44 countries. By understanding how these networks (organization-based pipelines) work in these peripheral economies, we highlight characteristics that may be typical of peripheral economies in general.

We find that peripheral economy inventors who collaborate with inventors in core economies tend to have more internationally dispersed networks. In addition, we provide some of the first empirical evidence on the Cantwell and Santangelo (1999; 2000) research on the dispersion of innovation activities involving tacit knowledge, in this case extending it to the context of peripheral economies.

The rest of the paper is structured as follows. Next, we review relevant literature. Then, we develop the theoretical bases of our analysis and derive our research hypotheses. Subsequently, data and empirical methods are described. Finally, we discuss our results and the associated implications.

Literature Review

Hypotheses

Our first hypothesis focuses on the relationship between the location of inventors and the level of disaggregation of patenting activity across national borders, specifically the dispersion of inventor networks. Inventors related to any country (e. g., Portugal) can be based locally (in this case, in Portugal) or based in foreign locations. In the second case, the relationship to Portugal arises from the fact that they work for an organization located in Portugal, i. e., an inventor working for a firm located in Portugal, but based in a foreign country. This foreign country can be a core economy or a peripheral economy.

In this paper, we focus on the case of inventors working for a firm in Portugal or Greece but located abroad. Organizations in peripheral economies (firms, research institutions, universities, etc.) seek knowledge from both local and non-local inventors, but they are likely to source the most complex, capability-driven, explorative knowledge (requiring the highest degree of collaboration) from core regions, since, on average, those locations have deeper pools of knowledge. Hence, the foreign-based inventors of peripheral-economy organizations who are located in core economies have access to wider innovation networks than those located in other peripheral economies. On the other hand, firms usually go to peripheral regions in search of exploitative, cost-driven knowledge. As the inventors they hire in peripheral economies undertake mainly exploitative work, they are only locally connected or at most, connected to a home economy subsidiary or to headquarters. Therefore, their networks will be more limited than those of inventors residing in core economies.

Drawing on the literature and the arguments discussed above, we state the following hypothesis:

Hypothesis 1: When foreign-based inventors of organizations based in a peripheral economy are located in core innovative economies, they will be connected to more internationally dispersed inventor networks than when they are located in other peripheral economies.

We analyze two aspects of the phenomenon of patenting activities in peripheral economies. H1 focuses on the dispersion of inventor networks of organizations based in peripheral economies. It examines the foreign-based inventors of these organizations, comparing those based in core innovative economies with those based in peripheral economies. On the other hand, the next hypothesis focuses on the type of knowledge. As discussed in the literature section, it is widely accepted that different activities within the value chain have different degrees of transferability, depending fundamentally on the extent of codifiability. More codifiable patenting activities can be either outsourced or disaggregated (even across national borders), through geographically dispersed innovation networks. In contrast, more tacit innovative activities, as a general rule, are more likely to be internalized and conducted by collocated teams. This is true in peripheral economies as much as in other contexts. Therefore, our second hypothesis is the following:

Hypothesis 2: When patenting in peripheral economies involves tacit knowledge activities, the inventor networks will be less internationally dispersed than when knowledge is more codifiable.

As Cantwell and Santangelo (1999) argue, there are two factors that facilitate the orchestration of tacit-knowledge patenting across dispersed networks. The first is organization-specific capability, typically associated with leading firms in the relevant knowledge space. The second is that the patenting is focused on competencies that are “non-core” for the company (Cantwell and Santangelo, 2000). Calantone and Stanko (2007) found that firms that are experienced in conducting exploratory research tend to outsource innovation activities (of any kind) to a higher degree. We argue that being an experienced innovator and having the capabilities associated with it will be most critical when the innovation is focused on tacit components. In addition, as argued by Cantwell and Santangelo (2000), for the largest and most experienced MNEs, most patenting with tacit components (such as design innovation) that is dispersed is typically not a core activity. Based on these arguments, we arrive at the following hypothesis.

Hypothesis 3: The relationship between tacit knowledge and the international dispersion of inventor networks will be moderated by the innovation capabilities of the firms. Leading innovative firms will be able to disperse their tacit knowledge patenting across borders to a higher degree than innovation laggards.

In summary, we hypothesize that in the context of peripheral economies, the dispersion of patenting inventor networks across national borders will depend on the combination of the location of inventors, knowledge tacitness, and organizational capabilities of the innovating organization. Our hypotheses only predict the spatial form of the co-patenting networks, based on the three factors listed above. We do not make predictions about the overall success of those networks, in terms of total output or in terms of the impact of the patents they generate.

The Empirical Context: Portugal and Greece

We chose two typical European peripheral countries as the empirical settings to illustrate the processes underlying patenting networks in peripheral economies: Portugal and Greece. Both countries can be considered textbook cases of European peripheral economies, as they display all characteristics usually attributed to such economies. These include the structure of production, the degree of internationalization and international openness, foreign subsidiary roles, linkages among actors, innovation-related indicators, connectivity with the core, and organizational/institutional characteristics (Benito and Narula, 2008; Molero, 1995; Molero, 1998).

Compared to core European Union (EU) economies, their economies are marked by a low degree of internationalization, low relevance of high-tech sectors and a low weight of high-tech exports. Their patent production represents only a minimal fraction of the European patenting activity (Paci and Usai, 2000). They also show a predominance of SMEs and micro-enterprises with low productivity and often offering non-tradable services (Simões and Godinho, 2011), and a paucity of indigenous MNEs, a relatively low supply of technology, and (in the case of Greece) a risk-averse national culture (Souitaris, 2001). Particularly in Greece, there is also a significant number of under-educated or under-qualified people in senior positions in numerous companies, which poses additional challenges to fostering an innovative culture (Souitaris, 2002).

At a more general level, both countries have a similar population size: 10.3 million for Portugal and 10.6 million for Greece (CIA, 2013b) and similar income levels: the GDP per capita (PPP) of Greece is US$27,300 and that of Portugal is US$32,200 (CIA, 2013a). They also have a comparable area and have the disadvantage of being located at the extremes of Europe, relatively far from the core economic and innovative regions in the continent.

As expected in peripheral economies, linkages among actors in these countries are modest. In Portugal, the low degree of autonomy of foreign subsidiaries limits linkages with the Portuguese science, technology, and innovation (STI) system (Tavares-Lehmann, 2008). Foreign-owned subsidiaries in Portugal also tend to source less locally than their domestic counterparts, since few local suppliers can fulfill the standards they require, in quantity and quality, though this is changing (Tavares-Lehmann, 2008). In Greece, there is also little engagement and interaction between the STI programs designed by the government and the innovative firms in the private sector, particularly MNEs (Collins and Pontikakis, 2006). Another problem in Greece is the uneven regional distribution of both big companies and R&D, with the bulk of activity concentrated in Southern Greece relatively little activity in other regions such as Thessaloniki (Huggins and Strakova, 2012).

Literature on patenting activities is more abundant for Portugal than for Greece. Most studies about Portugal (Godinho, 2009; Godinho et al., 2004; Godinho et al., 2008) show that the country is well below the OECD average in terms of patent indicators. Yet, there has been an acceleration in patent applications since 2000 (Godinho, 2009). The recent increase in international patenting is mainly driven by the business sector. Subsidiaries of foreign MNEs and born-globals have been particularly active in filing patents internationally, notably in the United States Patents and Trademark Office (USPTO) (Godinho et al., 2008). For high-tech firms, most of which are SME startups, patenting in the USPTO is a matter of reputation and “signaling” to potential partners and clients. MNE subsidiaries tend to centralize patenting processes, including patent applications, at headquarters or at a central R&D base. In Greece, there have been a number of programs (EPET I and II, STRIDE Hellas) aimed at increasing the scientific and innovative production of the country. In spite of steady increases in overall production of patents and publications since the 1980s, the country is still a clear innovation laggard in the context of the European Union (Collins and Pontikakis, 2006).

Data and Methodology

Estimation

Table 1 presents the summary statistics for our sample. The dependent variable is bounded, with a minimum value of 0 when all the inventors are in the same country, and a maximum observed value of 0.800. Of the patents in the data set, 50.8 % only have one inventor-country, which means there was no international collaboration involved. The other 49.2 % of the patents involved networks of collaboration between inventors in different countries. There is a large dispersion of innovative capabilities among the sample firms, as measured by their patent pool. The median firm in our sample holds approximately 40 patents.

We employ a multiple regression approach to test our hypotheses. As described previously, our dependent variable is double censored; the most appropriate technique for this type of dependent variable is a Tobit regression (GREENE, 2000: 905–926). Tobit models have been used in many studies with similarly censored dependent variables (Jeong and Weiner, 2012; Laursen and Salter, 2006; Mudambi and Helper, 1998; Ragozzino and Reuer, 2011).

Table 1:

Descriptive Statistics

Table 2:

Pearson Correlation Coefficients

Multicollinearity diagnostic checks were performed by running each model with an OLS regression and calculating variance inflation factors (VIFs). All the estimates showed values of less than 3, well below the commonly accepted threshold of 10 for VIF values (Chatterjee and Price, 1991). Finally, we acknowledge that there may be other factors not included in our model that affect both location and inventor dispersion. For this reason, we do not take our coefficients as indicators of causality but rather as indicators of associations between constructs.

Results

We ran three regression models to test our hypotheses (see Table 3). All models use censored Tobit analysis and the dependent variable is the dispersion of inventors across countries (measured for each focal patent).

Model 1 is the base model and Model 2 is our full model containing the interaction term Tacit knowledge activities by innovation leaders (LEAD_X_DES). Model 3 is similar to Model 2 but only includes patents linked to Greece. As predicted by our Hypotheses 1, higher GDP per capita is associated with more international dispersion of inventors. This implies that inventors located in core economies have access to richer networks of innovation. This finding is consistent with our conception. As we hypothesize, inventors of in richer economies usually can tap into deeper pools of knowledge, which usually means more geographically dispersed inventor pools collaborating in the patents. On the other hand, since firms in peripheral regions usually undertake exploitative, cost-driven knowledge creation, their connections tend to be more geographically restricted. The coefficients for Inventor-Country GDP per capita are significant but very small. Since our DV is an index, it is not possible to interpret the magnitude of the effect, and we acknowledge this is a limitation.

Hypothesis 2 focuses on tacit knowledge activities operationalized by design patents. We predict that design patents will be usually authored by co-located teams, due to the high component of tacit knowledge they contain. The degree of codifiability is one of the main determinants of the ability to transfer knowledge across distances. Codifiable knowledge can be exchanged across long distances, facilitating the dispersion of patenting teams. More tacit innovative activities, on the other hand, require co-location, observation, and learning-by-doing for their transmission, therefore setting a constraint for the dispersion of inventors. Therefore, when the knowledge is highly tacit, teams are more likely to be collocated than dispersed. This is true in peripheral economies as much as in other contexts. In other words, the geographical dispersion of teams involved in design patents will be less than for utility patents. The coefficients for DESIGN are negative in all models and significant in our full models 2. This is consistent with H2.

Finally, Hypothesis 3 predicts that innovation leaders are more capable to integrate tacit knowledge innovation across geographic space. This is based on Cantwell and Santangelo (1999, 2000) and Calantone and Stanko (2007), who predict that experienced innovators develop both the absorptive capacity, and the recombination and knowledge management capabilities to disperse even those knowledge creation activities that involve tacit knowledge. Less capable or experienced firms are unlikely to possess these competencies. In other words, innovation in design can only be geographically dispersed when it is carried out by leading firms. Our interaction coefficient LEAD_X_DES is positive and significant in both model 2 and 3, consistent with H3.

In terms of controls, MNE shows positive and significant coefficients. This is consistent with the notion that MNEs will have access to networks in multiple countries, which local firms will not be able to match. The geographical dispersion of assignees (ASSI_DISP) is positive and significant. This is intuitive; if a patent is coauthored by assignees dispersed in different countries, the inventors are also likely to be geographically dispersed. The coefficient for number of inventors (NUM_INV) is also positive and significant. This is not surprising either; the larger the group of inventors participating in the patent, the larger the chance that one or more of them are located in a different country. The coefficient for other organizations (OTHER_ORG) is also positive and significant. This is consistent with the notion that person-based linkages (the type favored by research institutions or universities) are easier to establish than organization-based linkages (the type favored by business firms). Research institutions or universities also facilitate the temporary co-location mechanisms (in the form of conferences, seminars, etc.) that are the basis for the creation of long-term person-based linkages. Finally, the coefficient for the Greek national system of innovation (GREE_NSI) is not significant, meaning that Greek innovators and their Portuguese counterparts do not show significantly different levels of dispersion.

To test the robustness of our data, we analyze data from other patent sources (EPO) and from other peripheral economies (Czech Republic, Slovakia and Slovenia). The data is not fully comparable, since some variables were missing. However, results (not reported here) seem consistent with our first hypothesis, that inventors in core economies are connected to more internationally dispersed inventor networks.

Concluding Remarks and Implications

The traditional development economics literature distinguishes between developed and developing countries (Meier and Rauch, 2005). Later literature identifies some members of the old developing country group that experienced rapid catch up along a number of dimensions as ‘emerging economies’ (Awate et al., 2012). But, with few exceptions, the growing diversity within the developed country group has not received much attention (Benito and Narula, 2008; Narula and Guimón, 2010). This paper focuses on the sub-group of developed countries that have been labeled ‘peripheral’ due to their relatively lower connectivity with the global economic system, as compared to the ‘core’ developed countries.

Table 3:

Tobit regression results for international dispersion of inventor networks

We use the comprehensive population data set of U.S. patents issued between 1976 and 2013 to Portuguese and Greek assignees (organizations) and inventors (individuals) to analyze the dispersion of inventor networks across national borders in these peripheral economies. Most studies of innovation systems are either couched at the level of organizations or at the level of individual inventors. We build on prior work on inventor networks (Balconi et al., 2004; Fleming and Marx, 2006; Zucker and Darby, 1996) and disentangle three factors that are associated with the dispersion of those networks: the location of the inventors, the type of knowledge, and the capabilities of the firm.

The first part analyzes the association between location of inventors and the international dispersion of inventor networks. Our findings are consistent with our conception that inventors located in core innovative countries have access to more internationally dispersed inventor networks. Thus, interaction with them will provide organizations based in peripheral economies with the potential benefits derived from this dispersion. In contrast, too much reliance on local knowledge sources may be harmful for innovative capacity and can lead to a “technology trap” (Gewin, 2005).

The second part explores how the tacitness of the knowledge involved in the innovation process hinders dispersion. Consistent with our conception, we find that design patents are associated with less dispersed inventor networks. This relationship, however, is moderated by the capabilities of the firms conducting the innovation. Highly innovative firms develop capabilities that allow them to conduct this type of innovation in a more dispersed manner. These findings are consistent with our second and third hypotheses. To the best of our knowledge, this is the first empirical testing of the theoretical work of Cantwell and Santangelo (1999; 2000) about the factors affecting the dispersion of tacit knowledge creation.

We believe our work has two types of implications. For academics, it opens the way to the exploration of a potentially very interesting area of inquiry: the characteristics of innovation in peripheral economies and the differences between the creation of tacit and codified knowledge in those contexts. Further work will be needed to disentangle the complex realities of these economies, but we think this a first step in that direction.

For policymakers, we provide some important distinctions about the factor that may affect connectivity in peripheral economies. For economies that are striving to catch up with the core, understanding these drivers may prove to be a very valuable tool. Concerning policy, the way to diminish the disadvantages of peripherality is to increase connectivity – by promoting the presence of locally based (domestic and foreign-owned) actors in international innovation and supply networks. Such connectivity to global value chains is a key aspect of high levels of local value creation (Humphrey and Schmitz, 2002; Mudambi, 2008; McWilliam et al., 2020).

In this context, our findings highlight the crucial role of the individual level of analysis (networks of inventors). Such connectivity requires a strengthening of system linkages (Mudambi, 2008) and “systemic density” (Godinho and Simões, 2013). Our results suggest that connectivity to core economies would potentially yield the highest benefits for peripheral economies, and this is where policymakers should focus their efforts. A context where technologies reduce spatial transaction costs, but policies may increase them, tends to favor large MNEs, because they have the resources to afford these transaction costs. It is therefore important that policies interfere in connectivity as little as possible, to create a level playing field for all companies, large and small. Given that linkages and networks need time to develop, consistency and predictability of policies are key factors.