Contribution to the international scientific production

Our first indicator of internationalisation of national scholarly communities is the contribution given by scholars, affiliated to institutions based in the respective countries, noting the overall number of articles published by the selected international journals.Footnote ⁴ For simplicity, and with the caveats discussed above in mind, let us call this a measure of “international scientific production”. This is shown in Fig. 1. The left quadrant includes the world’s top-20 countries for scientific production, while the European top-20 are shown in the right quadrant. The grey bars display figures for the decade 2000–2009, while black bars are for 2010–2019. Additional indicators are available in Tables 5 and 6 in the Appendix.

Fig. 1

Contribution to international scientific production. World's top-20 (left quadrant) and European top-20 countries (right quadrant)

The scientific production of political scientists is highly concentrated in relatively few countries. Indeed in both quadrants, the y-axis is scaled down (truncated) to make the figures for smaller national communities visible; the data for the US and the UK are displayed in a call-out. In the decade 2010–2019, US (41%) and British (15%) institutions together hosted more than half of the (co-)authors of the articles published in the top-100 journals. Just ten countries hosted the authors of 80% of all articles. European countries were largely represented in this highly productive core: seven among the top-10 for scientific production and fourteen among the top-20 were European countries.

Within the general picture of a scientific production concentrated in a relatively small number of countries, however, the data also reveal a tendency towards diversification. In the previous decade (2000–2009), more than 70% of the scholars who published in the top-100 journals were based in the US or the UK. The top-10 countries hosted 87%, and the top-20 hosted 95%, of all authors. Over time, the international scientific production became relatively less concentrated across countries.Footnote ⁵

European countries were key to the increased diversification of international scientific production, in two ways. On the one hand, the European scientific production increased markedly. On the other hand, the contributions to the international scientific production became much more evenly distributed among European countries themselves (see Table 5). To be sure, even in 2010–2019 the UK still provided an affiliation to a large share—roughly one third—of European contributors to publications in the top-100 political science journals; but down from 45% of Europe’s contributions in the previous decade. And while in the second decade, the British contribution more than doubled in absolute numbers (from less than 5000 to more than 10,000), so did the scientific production of the 20 more productive European countries. Meanwhile, contributions by scholars based in Germany, Ireland, Italy, Austria and Hungary more than tripled.Footnote ⁶

Clearly, UK-based institutions still provide a large share of the European contributions to publications in leading international political science journals. Yet, the increase in the amount of contributions from + other European countries has been remarkable.

Participation in international collaborative publications

Our second indicator of internationalisation of national scholarships is the contribution to collaborative studies co-authored by scholars based in different countries. Multiple country publications (MCP) are shown in Fig. 2, this time only for the 20 European countries with the higher contribution to the international scientific production. For each country, the bars indicate the absolute number of MCP; alongside each bar, the share of MCP on the total number of publications contributed by that country is included. Separate bars for each country show figures for the first and second decades under analysis.

Fig. 2

Involvement in collaborative work by country, European top-20

The generalised tendency towards an increased involvement in international collaborative publications outpaced the broader growth of the global scientific production. If we consider the top-35 countries for scientific production (which include the European top-20), single-country publications (SCP) increased by 1.57 (about 150%) between decades; meanwhile, MCP increased by 5.56 (or about 550%) on the average.

This is better appreciated by observing the performance of single countries. In the decade 2000–2009, MCP represented 20% or more of the total scientific production in seven European countries: the Netherlands (20%), Belgium (22%), Portugal (26%), Poland (23%), Romania (25%), Ukraine (20%) and Bulgaria (33%). Except for Belgium and the Netherlands, these were relatively small political science communities producing a limited number of publications. The European average was 17%. In the following decade, MCP grew to 27% of the total production on the average. They were 20% or more of the total scientific production in the large majority of European countries, with just eight exceptions.Footnote ⁷ In relative terms, MCP had increased massively in most European countries. In 2010–2019, they were between four and five times as numerous as in the previous decade in Italy, Switzerland, France and Finland; and between five and eight times more numerous in Denmark, Spain, Sweden, Austria, Portugal and Hungary.

International collaborative publications seem to have been an important component of the general growth of the international scientific production between the two decades. This was a generalised tendency in the large majority of national political science communities. It was certainly so in European countries, where the change was indeed even more marked. These figures complement the description of a growing presence of Europe-based political scientists in the international political science scholarship, with a broader and more diverse geographic basis and a parallel massive increase in collaborative publications. International collaborative research seems to be establishing itself as a key engine for scientific production, in Europe as well as globally. The next section discusses how international collaboration works in practice, and how it has changed along the two decades observed.

Country collaboration network analysis

Social network analysis permits to look closely into actual patterns of international cooperation. Single countries can be represented as nodes in a network of country collaborations, as in Figs. 3 and 4. Two countries are connected if at least one article was co-authored by scholars working in institutions based in those countries. The higher the number of co-authorships between scholars based in two countries, the stronger the connection between those countries (and the larger the edge joining the related nodes in the network).

Fig. 3

Country collaboration network, 2000–2009. Entire network in the inset, and fifty most connected countries zoomed in in the chart

Fig. 4

Country collaboration network, 2010–2019. Entire network in the inset, and fifty most connected countries zoomed in in the chart

Several indicators of connectedness and centrality can be used to obtain a deeper understanding of the network configuration. Table 1 presents some aggregate network-level indicators describing the internal structure of the network and its development over time. Tables 2 and 3 instead provide core 8network statistics at the level of single countries for the 50 nodes with the higher degree (most connections) in the network. Statistics about all countries are shown in Tables 10 and 11in the Appendix that also include some additional indicators of centrality.

Table 1

Network connectedness and development

Table 2

Country collaboration network. Clustering and centrality of single countries within the network. Fifty most connected countries, 2000–2009

Table 3

Country collaboration network. Clustering and centrality of single countries within the network. Fifty most connected countries, 2010–2019

Structure and evolution of the network

Aggregate network indicators, shown in Table 1 separately for the two decades, consistently point to a broadening and tightening of the network. While it expanded from 106 to 129 countries from the first to the second decade, the network also became twice as densely connected. Despite the larger network size, communication flows (that is, connections between scholars based in different countries) did not become more difficult; the average length of the shortest path connecting two countries actually decreased slightly.

Transitivity measures the extent to which the relations connecting two nodes in the network are “passed on” to other nodes (Wasserman and Faust Reference Wasserman and Faust1994: 243): how probable is it that if, say, scholars in the US are connected to scholars in Estonia, and scholars in Estonia are connected to scholars in Lithuania, then scholars in Lithuania are also connected to scholars in the US?Footnote ⁸ This indicator also increased markedly from the first to the second decade, which points to the development of more inclusive clusters in the second decade.

Finally, the network became slightly more centralised. Degree centralisation equals one when all nodes connect to a single one, and not to each other (as in a star); it is zero when all nodes are equal (as in a circle). Higher degree centralisation therefore means that there are relatively less very central nodes. This development is consistent with network theory, which sees centralisation as an indicator of network efficiency and posits that while networks grow in size, they will also become more centralised.Footnote ⁹

What are these best-connected countries? How did the structure of the collaboration network change over time, and what did it mean for the general organisation of international collaborations? To address these questions, we now turn to the descriptive analysis of the network.

2000–2009

Let us begin with the network of international collaboration in the period 2000–2009. For ease of visualisation, Fig. 3 zooms in on the more connected core comprising 50 countries; the entire network is shown in the inset. Nodes (countries) are coloured according to the cluster in which they are placed.Footnote ¹⁰ Table 2 complements the graphical visualisation of the network with several centrality indices for single nodes: the cluster to which countries belong; their degree (number of connections); normalised closeness (the steps required to access every other node); betweenness (brokerage potential, measured as the normalised number of shortest paths crossing each node); and authority (connection to nodes that link many other nodes).

In the first decade, the network of 106 countries hosting scholars who published in the political science journals selected for this study included three main clusters, comprising overall 57 countries. The remaining 49 countries were isolated nodes in the network (as can be seen in the inset of Fig. 3). The main cluster (red in the figure) included 41 countries and had the US—the country scoring highest on all network indicators by far—at the centre. We may call this the global cluster. The UK was the second most central actor in this cluster (and in the broader network), alongside of other countries to which it is historically connected, such as Canada, Australia, New Zealand—as well as Ireland in the first place. This cluster has a few more well-connected countries, notably Israel, China, and Japan.

Importantly, 12 European countries were part of this cluster. In addition to Ireland and the UK, these were Cyprus, Estonia, Greece, Iceland, Latvia, Lithuania, Poland, Portugal, Romania, Russia, Slovakia and Ukraine. Of these, Greece, Portugal and to some extent Estonia had a relatively high degree compared to other countries within the cluster, but limited network centrality and virtually no brokering potential. The other European countries were part of this collaboration cluster primarily because scholars based there had collaborated with colleagues from the main hubs within the cluster.

Most remaining countries within the global cluster were weakly connected in all respects. These included the larger Latin American countries (Argentina, Brazil, Colombia, Chile, Mexico, Venezuela), as well as several African (Burkina Faso, Egypt, Ethiopia, South Africa, Zimbabwe,), Asian (India, Indonesia, Korea, Kyrgyzstan, Philippines, Singapore, Sri Lanka) and Pacific (Fiji) countries. Such countries had a low degree (number of connections) and no brokering capacity; in general, they scored poorly on all connection indices. Overall, they were peripheral to the global cluster and were part of it because of their collaborations with few countries from its core.

Most European countries were part of the two remaining clusters. One larger European cluster included 12 nodes, with Germany and the Netherlands at the centre. They clearly had a key brokering role within the European cluster, being pivotal to collaborations among scholars from many other European countries; but their degree and centrality scores made them actually quite central to the entire network. To a relatively smaller extent, this was also the case for Italy and Switzerland, and less so for Spain and France.

Belgium was a somehow special country within the European cluster. Its betweenness was extremely high compared to its degree (still relatively high compared to the size of the country). Although limited in mass, scholars based in Belgium were able to place themselves at the centre of a tight web of international collaborations, particularly with other European colleagues. Austria and Hungary were the remaining European countries with a marked degree and some betweenness in the network. The Czech Republic, Slovenia and Luxembourg had some degree but no brokering capacity.

The second cluster of European countries was composed of the Scandinavian countries: Denmark, Finland, Norway and Sweden. Their relatively high degree, as well as the web of collaboration within and outside the cluster, is what placed these countries in a cluster of their own. Among them, Norway, Sweden and Denmark had the higher degree (comparable to such countries as Italy or Switzerland in the main European cluster, or Ireland and Israel in the global cluster) and also a significant centrality. Sweden was second by degree, but it was by far the main hub, both within the cluster and with other countries.

Finally, the network included a set of 49 individual nodes with no systematic relations with any cluster of countries. Scholars based in these countries either did not publish collaborative work with colleagues based in other countries, or (in a smaller number of instances) their collaborations happened mainly or exclusively with single countries. In short, they were not systematically connected to international networks of collaboration for articles published in the top-100 international journals. This set of countries, that constituted about half of the entire network, included eight European countries: Malta, Bulgaria, Serbia, Croatia, Belarus, Moldova, Monaco and Montenegro (in decreasing order of network degree). None of them had any significant centrality or brokering capacity within the overall network.

2010–2019

Figure 4 displays the network of country collaborations in the decade 2010–2019. Again, the figure zooms in on the top-50 countries for degree within the network, while the entire network is shown in the inset; single countries are coloured based on the subgroup to which they belong. Indicators of centrality and connectedness of single countries are shown in Table 3 for the 50 most connected countries within the network. Table 8in the Appendix provides centrality indicators for all countries in the network.

By 2010–2019, the network had expanded in size, but the isolated nodes had decreased both in relative and absolute terms: less than one third of the countries—37 out of 129 countries—were now excluded from collaboration clusters, compared to almost half of the 106 countries composing the network in the previous decade. International collaborations were still organised around three clusters of scholarly communities, and these groupings maintained their overall identity: one global cluster and two (mainly) European clusters. However, all expanded in size and deepened in terms of interactions.

The global cluster expanded to include 62 nodes. The US and UK maintained their centrality within the cluster (as well as within the broader network): their score on degree and centrality measures actually increased markedly in the second decade. Canada and Australia followed, also part of the top-10 most connected, central, and in-between countries in the broader network. More countries within the global cluster became tightly connected (see Table 3). Canada, Australia, Israel, China, Ireland and Turkey all had a degree (number of international co-authorships) higher than 1000; Russia, South Africa, Korea, Japan, Singapore, Brazil, Mexico and New Zealand higher than 500.

Although significantly lower in network degree, Ireland and Turkey stood out for their high centrality score and brokering capacity within the network (as shown by the betweenness and hub indicators). The latter in particular was rather well connected to many countries in the European cluster, as was the case for some Latin American countries, such as Brazil and Chile. Korea, Japan and Singapore also were quite well connected within the global cluster; less so with countries in the European cluster.

The global cluster included ten European countries (down from fourteen in the previous decade), four of which were European Union members. In addition to the UK and Ireland, European countries in the global cluster were Russia, Cyprus, Estonia, Ukraine, Serbia, Iceland, Croatia, and Belarus. Russian scholars stood out for the scope of their connections with Europe, having co-published with scholars from 18 (other) European countries — a massive increase from the previous decade, when Russian scholars only published collaborative work with scholars from the UK, Germany and Finland. With the partial exception of Cyprus and Estonia, the remaining European countries within the global cluster had limited connections within, and poor or no connection outside of it.

By 2010–2019, most of the scholars from European countries had tightened their collaboration with European colleagues. As a result, the larger European cluster grew to include 20 nodes; up from 12 in the previous decade, and now also comprising three non-European countries (Qatar, Jordan and Venezuela). The top-three countries for connection, authority and centrality within the main European cluster—Germany, Switzerland and Italy—were also among the top-10 most connected countries of the broader network of international collaborations. France was among the top-10 for most centrality indicators, although not for degree. In general, most countries within the European cluster were well connected within and outside of it; relatively less so were scholars in Slovenia, Lithuania and Bulgaria, as well as the three non-European countries.

The second European cluster also grew in size, from four to nine nodes. The four Scandinavian countries were joined by Belgium, Luxembourg, and the Netherlands—another well-connected and internationalised community of scholars—as well as Malta and Mozambique. While no longer a Scandinavian cluster, this was still a Northern European one. Malta was part of it because of some repeated collaborations between a few productive scholars with colleagues in the Netherlands, and Mozambique because of co-authored publications with scholars in Norway and Denmark.Footnote ¹¹

The Northern European cluster was tightly connected internally as well as externally. Scholars based in the countries that were part of this cluster used to collaborate systematically with colleagues from other countries of the same group. However, this was also a highly internationalised cluster—and a very successful one in terms of scientific production. Three countries—the Netherlands, Sweden, and Denmark—ranked among the world’s top-10 for degree and on most other indicators of centrality and authority. Belgium and Norway followed shortly after.

The Netherlands can be considered an exceptionally successful case. Despite its small size, it ranked fourth globally on most indicators; after the US, the UK and Germany but before Italy, Canada, Australia, or France. Although to a slightly smaller degree, Switzerland, Sweden, Denmark, Belgium and Norway are comparable cases of relatively small but extremely productive and internationalised political science communities. All of them, except Switzerland, were part of the Northern European cluster.