Introduction

Innovation systems [1,2,3] and regional science [4] are two research fields that relate to spatial economics and economic geography from different approaches. Regional science focuses on the locational dimension of human activities in the context of their institutional structure and coordination. It is only implicitly aimed at the study of innovation processes. In contrast, innovation systems research studies the economic actors and institutions that contribute to generating innovation by focusing on innovation processes [5].

According to Forrest [6], firms are open systems that receive feedback from their external context. Generally, the origin of innovations in a firm or an organization, in a country or a region, will depend not only on its human capital and internal factors but also on external factors such as the institutions and agents of each country or region. These interdependencies between the firm and its context give rise to the innovation systems, which comprise a great variety of institutions, networks, and interrelationships.

The application of the innovation systems to a subnational area (regional or even local) was carried out to mitigate the most severe problems related to the national scale [7]. This regional approach was dubbed regional innovation systems (RISs) by Philip Cooke in 1992 [8]. Research on regional innovation has grown significantly since the term “RIS” was coined. This growth owes to the greater intensity of international competition in the globalized economy, the shortcomings of traditional regional development models and policies, and the emergence of successful clusters of firms and industries in many regions around the world [9, 10]. The literature on RIS provides extensive descriptions and analysis of relationships between innovation, learning, and economic performance of particular regions [11,12,13,14].

In view of this background, the main objective of this paper is to identify the research trends that have characterized RIS research from its beginnings to the present day. Bibliometric techniques are applied to data on all RIS-related publications indexed in the Web of Science Core Collection (WoS CC) database to outline the RIS conceptual framework and select the most influential RIS articles. The aims, scope, and conclusions of the most influential articles are then studied to identify the main research trends.

The rest of the article is structured as follows. The next section describes the theoretical framework; then there is a section that explains the methodology and another section that presents the results of the bibliometric analysis and of the review of the most influential articles. Finally, the last section provides the main conclusions.

Theoretical framework

The systemic approach to innovation emerged in the context of debates over industry policy in Europe at the end of the 1980s, giving place to the national innovation systems (NIS). The collaboration between Chris Freeman, Richard Nelson, and Bengt-Åke Lundvall in the International Federation of Institutes for Advanced Study (IFIAS) was crucial for the subsequent development of the concept. Three books pioneered the idea of the NIS: Technology policy and economic performance: lessons from Japan, by Freeman [1], National systems of innovation: towards a theory of innovation and interactive learning, edited by Lundvall [2], and National innovation system: a comparative analysis, edited by Nelson [3].

According to these pioneers, a NIS consists of a network of economic agents together with the institutions and policies that influence these agents’ innovation behavior and performance. Within the NIS-based conceptual framework, innovation is an interactive process in which firms that interact with and receive support from institutions and organizations (e.g., industry associations; R&D, innovation, and productivity centers; standard-setting bodies and institutes; and universities and training centers) play a key role in bringing new products, new processes, and new forms of organization into economic use [15].

Contrary to what might seem in a globalized world, the national and regional approaches are key factors to build the relational networks that firms need to innovate, so the role played by nations regarding innovation has become even more important [16]. For this reason, public administrations are investing heavily in stimulating innovation processes, as well as in improving the business environment through the design, implementation, and evaluation of innovation policies [17,18,19]. The innovation systems approach is widespread in Scandinavia and Western Europe in both academic and policymaking contexts. Not only national public administrations but also supranational organizations such as the Organisation for Economic Cooperation and Development (OECD), the European Union (EU), and the World Bank have adopted the NIS approach to develop innovation policies [20, 21].

The systemic approach to innovation may be complemented with other two models: (1) on the one hand, the Triple Helix [22] establishes that the conditions for the innovation process are mainly determined from the relationship between three kinds of agents: academia (science or universities), industry (or firms), and state (government or public sector); (2) on the other hand, the open innovation [23] reinforces the previous models because it considers the firms’ boundaries permeable to the exterior, so they can be affected by the external context.

The RIS may be taken into account as a special case of the NIS, arising from the application of a subnational focus to this concept and seeking to reduce the complexity by considering a smaller area. Moreover, while national, international, and sector factors are essential, the regional dimension is also crucial. Therefore, different innovation policies should be applied to different types of regions to deal with particular innovation barriers affecting a given region [24].

The theoretical foundations of the RIS lie in the regional scaling of economic processes and in systemic and evolutionary approaches to innovation and learning. The innovation systems approach considers innovation a systemic and dynamic process that emerges from interactive learning among firms and other organizations such as universities and research centers [25]. Accordingly, subsequent development of the RIS concept has linked research on regional science to evolutionary economics and the evolutionary dynamics of change and adaptation of regions [26, 27], the economics of innovation [28, 29], theories of interactive learning [2], and institutional economics [30].

Regardless of whether the focus is on the institutional and organizational dimensions [7, 30], the systemic approach of innovation [16, 31], or the evolutionary perspective [5, 32, 33], the broad consensus is that innovation policies must consider each region’s specificities, which depend on political, economic, and sociocultural factors, as well as the legal, technological, and environmental context [14, 17, 34, 35].

Based on this theoretical framework and focusing on the RIS evolution and possible prospects, the central research question of this study was: what have the main underlying RIS research trends been over time?

Method

Bibliometric analysis

Bibliometrics [36] refers to the quantitative analytical methods used to analyze citations by articles in academic journals. The goal of bibliometrics is to evaluate the impact of publications based on the extent of their dissemination [37]. The most commonly used bibliometric indicators are the total number of publications, the total number of citations, and h-index [38]. The h-index is a popular indicator among researchers because it combines the number of publications and citations into a single indicator. For example, if an author has an h-index of N, this means that that author has written N documents that have been cited at least N times [39]. The ratio of number of citations received by an article per year (as opposed to total citations of that article) might be the preferred measure to identify the most influential studies, because by only considering the total number of citations, we create a bias toward older documents.

Bibliometric maps are graphical representations of how research fields and topics and individual papers are interrelated [40]. These tools map a research field, helping researchers identify its cognitive structure, evolution, and main actors and providing a clear visualization of results [41]. Analysis of keyword co-occurrence is based on the study of the most common keywords in documents. Thus, a bibliometric map of keyword co-occurrence helps identify the conceptual framework of a research field [42]. Other common bibliometric maps include bibliographic coupling, co-citation, and co-authorship. Bibliographic coupling and co-citation map the relationship between key ideas in a specific scientific domain. Bibliographic coupling [43] occurs when two documents cite the same third document (the number of references shared by citing documents), while co-citation [44] refers to situations where two documents are cited in one or more published articles (the number of times they are cited together). Finally, co-authorship analyzes the number of co-authored documents to study the social structure and research collaboration networks [45]. VOSviewer is used to produce the bibliometric maps presented in this paper [46], although other bibliometric software tools also exist. Each tool has certain advantages and disadvantages [47]. VOSviewer is freely available. Further information can be found at http://www.vosviewer.com/.

First, the bibliometric technique of keyword co-occurrence in the selected RIS research documents is used to perform descriptive analysis of the theoretical and conceptual framework of RIS. Then, the total number of citations and the ratio of number of citations per year are calculated for each document to identify the most influential RIS studies. Finally, the most influential studies are reviewed to determine the main trends in RIS research from its beginnings to the present day.

Data

All data were taken from the Web of Science Core Collection (WoS CC) database. This database belongs to Clarivate Analytics. WoS CC is internationally recognized by researchers as a high-quality source of information for searching and evaluating different types of publications and journals [48].

The search implemented in WoS CC to extract the data for this paper was Topic = “regional innovation system” OR “regional innovation systems” OR “regional innovations system” OR “regional innovations systems” OR “regional system of innovation” OR “regional systems of innovation” OR “regional system of innovations” OR “regional systems of innovations”. This search was conducted in April 2019. All years up to and including 2017 were considered. The search returned 972 studies. Studies exclusively classified as proceedings papers were excluded because most had received 0 citations and were therefore irrelevant to the analysis. Thus, the final set of studies comprised 680 studies. This set of studies consisted of 533 articles, 63 book chapters, 29 proceedings papers, 18 reviews, 18 book reviews, 11 editorials, four books, and four editorial book chapters. Note that WoS allows the same study to be classified as several document types.

This set of 680 studies had received 16,166 citations by the end of 2018, with a ratio of 23.8 citations per study and an h-index of 60. This h-index indicates that 60 of these publications had received at least 60 citations. These 680 studies cover 39 research areas, although there are just five research areas including more than 100 studies: Business & Economics (364), Public Administration (242), Environmental Sciences & Ecology (220), Geography (218), and Urban Studies (131). Note that the same study can cover multiple research areas.

Figure 1 shows the yearly evolution of the total number of publications and the total number of citations. The first RIS research study indexed in WoS CC was published in 1992. “Regional Innovation Systems: Competitive Regulation in the New Europe” [8] is widely accepted as the paper that coined the term “RIS.” Since then, documents have been published every year, except 1993 and 1996. The number of studies each year has oscillated, reaching 23, 50, and 70 studies in 2005, 2010, and 2015, respectively. Although the number of RIS studies increased significantly in 2010 and 2011, with 50 and 64 publications, respectively, a consistent upward trend of annual publications cannot be observed until 2015. The maximum number of studies in a given year was 84 (in 2017).

Fig. 1
figure 1

Total number of publications and citations per year

The annual evolution of the total number of citations increased steadily from 1999 onward, with the exception of 2013. The number of citations decreased from 1179 in 2012 to 1172 in 2013. The 100-, 500-, 1000-, 1500- and 2000-citation thresholds were reached in 2004, 2009, 2012, 2015, and 2017. The maximum number of citations (2054) occurred in 2017. The evolution of the total number of publications and citations per year reflects the scientific community’s growing attention and interest in RIS research.

Results

Conceptual framework: common keywords and co-occurrences

According to Callon et al. [42], analysis of the co-occurrence of keywords is used to study the conceptual structure of a research field. Figure 2 maps the keyword co-occurrence for the most common keywords in RIS research from 1992 to 2017. The map is based on a threshold of 20 occurrences and the 100 most representative links. Table 1 presents the 45 keywords with at least 20 occurrences.

Fig. 2
figure 2

Map of keyword co-occurrence in RIS research (1992–2017)

Table 1 Most common keywords

The concepts captured by the keywords are diverse. Knowledge, innovation, clusters, policy, networks, systems, R&D, industry, and firms appear to be the most frequently used keywords in RIS research, with more than 80 occurrences each.

The VOSviewer keyword co-occurrence map shows four clusters. Clusters generated by VOSviewer are for guidance and help identify the most connected keywords according to the co-occurrence between them [46]. The principal keywords in these clusters (1–4) are the following: (1) knowledge, innovation, policy, and systems; (2) clusters, networks, and industry; (3) R&D, firms, and performance; and (4) technology and innovation system.

Research trends: the most influential RIS articles indexed in WoS CC

Many influential papers on RIS have been published. One method to identify these influential papers is to classify publications based on the number of citations. The number of citations reflects the influence and popularity of the article and the attention it has received from the scientific community [48]. The ratio of number of citations per year was also calculated for all publications because the total number of citations has a certain bias toward older papers that have had longer to accumulate citations.

Table 2 presents the most influential RIS articles indexed in WoS CC based on the total number of citations. To determine the most influential articles, we used two criteria: (1) articles that have received at least 150 citations and (2) articles with a minimum ratio of 13 citations per year.

Table 2 Most influential studies in RIS research

Table 2 displays 12 articles with at least 150 citations and 14 articles with a minimum ratio of 13 citations per year, resulting in a total of 17 articles. This table includes the ranking by total citations (RTC), the total number of citations (TC), the ratio citations per year (C/Y), and the ranking by citations per year (RCY). Nine of the articles in Table 2 meet both criteria (at least 150 received citations and at least a ratio of 13 citations per year). Surprisingly, the article with the highest ratio of citations per year, focused on entrepreneurial ecosystems, was published recently (in 2017) and has received 106 citations in just 2 years (2017 and 2018).

Finally, RIS research trends are identified by reviewing the most influential articles in this field. This review was based on the aims and scope, and conclusions of the articles (see Table 2). On the basis of the review of the articles’ aims, scope, and conclusions, Table 3 presents the main RIS-related topics dealt with by each article. The articles are ordered chronologically within each group. The article review reveals three groups of trends in RIS research.

Table 3 Main trends in RIS research

First, the innovation systems research trend comprises six articles. All were published in the 1990s, except the articles by Todtling and Trippl [24], which focuses on regional-based innovation policies, and Oh et al. [58], which is a critical examination of the fledgling concept of the innovation ecosystem. These articles are especially aimed at exploring the systemic approach of innovation from different perspectives such as institutions, organizations, networks, policies, regulations, or the evolutionary approach.

Second, the knowledge management research trend—knowledge management can be defined as the process of creating, sharing, using, and managing an organization’s knowledge and information [60]—comprises nine articles, all published in the 2000s, except the article by Yam et al. [56]. This research trend focuses on the importance of economically useful knowledge in the regional innovation process, including the study of knowledge creation, knowledge spillovers, knowledge diffusion, knowledge flows, knowledge bases, knowledge-intensive business services (KIBS), R&D, patents, and clusters [61,62,63].

The third trend is that of entrepreneurial ecosystems, which are “combinations of social, political, economic, and cultural elements within a region that support the development and growth of innovative start-ups and encourage nascent entrepreneurs and other actors to take the risks of starting, funding, and otherwise assisting high-risk ventures” (Spigel, p. 50) [57]. This research trend comprises two articles published recently (in 2017). The fact that only two articles have followed this research trend is probably because it is a recent research trend. Consequently, it requires more research and development. Entrepreneurship is a potential source of innovation that has become a popular topic in recent years because countries and regions must innovate and generate competitive advantages based on local agents, processes, and dynamics to compete in the globalized world economy [64, 65].

Co-citations between the most influential RIS articles

This section presents the results of analysis of citations to the 17 most influential RIS articles. The aim of this analysis is to identify the relationships between these articles based on the number of times they are cited together. A total of 4199 studies indexed in WoS CC have cited these 17 most influential articles.

Figure 3 shows the VOSviewer co-citation map for the most influential articles based on the citations appearing in the 4199 citing studies. This figure shows the name of only the first author of each document, in addition to the publication year and the journal where the article was published. The visibility of labels in VOSviewer is not optimized, so some labels may not be visible because of a lack of space [46]. However, Table 4 shows the publication year of each article, its authors, and the full title of the article. Table 4 presents the 17 most influential articles grouped by the clusters in VOSviewer to identify all nodes of Fig. 3, including the two nodes that have no associated text in the figure.

Fig. 3
figure 3

Co-citation mapping between the 17 most influential RIS articles

Table 4 Co-citation analysis between the 17 most influential RIS articles

Table 4 is ordered chronologically within each cluster and displays data for the following variables: total number of citations (TC) received by each article; total number of co-citations (TCo), which counts the times that each article has been cited with any other article; number of co-citations between the 17 most influential articles (Co), which counts the times that each article has been cited with any of the other 16 most influential RIS articles; number of co-citation links of each article with the other 16 most influential RIS articles.

The number of co-citations and links between the 17 most influential RIS articles can be used to identify how these articles are interconnected. Interestingly, the articles by Cooke et al. [7] and Asheim and Coenen [11] are the only articles that have co-citations with all the other most influential RIS articles. The articles by Hansen and Niedomysl [55] and Yam et al. [56] are the articles with the fewest links (six links). Articles with more citations are more likely to have more co-citations and links. Surprisingly, however, the most recent articles [57, 59] have a significant number of links (10 links).

Conclusions

This paper identifies the main research trends in the RIS literature using bibliometric analysis of RIS studies indexed in WoS CC. The search was conducted in April 2019 and considered all years up to and including 2017. The search returned 972 studies. Proceedings papers were discarded because most had received 0 citations. After these proceedings papers had been removed, 680 studies were left. This set of studies had received 16,166 citations by the end of 2018, with an h-index of 60. These values reflect the high influence, popularity, and impact of RIS research among academics. Business & Economics, Public Administration, Environmental Sciences & Ecology, Geography, and Urban Studies are the research areas with most published RIS studies. The literature review indicates that this field covers a diverse range of concepts. Based on this review and the bibliometric analysis, three underlying research trends can be identified.

First, the general conceptual framework of RIS was built by analyzing the most common keywords and their co-occurrences in the set of 680 studies. Knowledge, innovation, clusters, policy, networks, systems, R&D, firms, and industry were the key concepts.

Second, the most influential RIS articles were identified by calculating the total number of citations and the ratio of number of citations per year of each article. The criteria of having at least 150 citations or a minimum ratio of 13 citations per year were then applied to identify the 17 most influential articles published between the years 1992 and 2017.

Third, the review of these articles reveals three main research trends that have dominated RIS research since its beginnings to the present day. The first is innovation systems research, which was mainly conducted in the 1990s. Innovation systems research focuses on the systemic approach of innovation in different contexts in terms of institutions, organizations, networks, policies, or regulations. The second is knowledge management research, which has been prominent since the beginning of the 2000s. This area includes knowledge creation, knowledge spillovers, knowledge flows, knowledge-intensive business services (KIBS), and different knowledge bases. It also considers other knowledge-related activities, processes, and agents such as R&D, patents, and clusters. The third is entrepreneurial ecosystem research, which has emerged in the last few years because of the key role of the social and economic context in local and regional entrepreneurship. The analysis of co-citations between the most influential RIS articles and the reviews of these articles shows that these research trends are strongly interconnected and are linked to other concepts such as R&D, firms, industry, innovation policy, clusters, patents, and technology. In addition, the high number of co-citations of the 17 most influential RIS articles and the significant numbers of co-citation links among them corroborate this finding.

Finally, this study has some limitations. First, RIS documents that are not indexed in WoS CC were not included in the set of studies. Use of a different database and the inclusion of proceedings might have affected the results. Second, although the co-citation analysis reveals a relationship between the co-cited articles, this relationship may not necessarily imply similarities between the studies. For instance, co-cited articles may be cited in two unrelated parts of the citing document. Although researchers should consider these limitations, this paper nonetheless sheds light on the field of RIS research.