Research Collaboration: Characterizing the global trends during the last five decades

Prashasti Singh^*, Vivek Kumar Singh^**, Hiran H. Lathabai^***, Aakash Singh^****

^*prashasti.singh8@gmail.com

0000-0003-0846-0450

School of Computer Science, University of Petroleum and Energy Studies, Dehradun, India.

^** vivek@bhu.ac.in, ^****aakash.singh10@bhu.ac.in

ORCID: 0000-0002-7348-6545, ORCID: 0000-0002-6213-718X

Department of Computer Science, Banaras Hindu University, Varanasi, India.

^*** hiranhl007@gmail.com

ORCID: 0000-0002-5633-9842

Amrita CREATE, Amrita Vishwa Vidyapeetham, Amritapuri-690525, Kerala, India.

Abstract: Collaborative research has become an integral part of any flourishing academic or industrial ecosystem. Collaboration is not only associated with higher research productivity but has also been found to be positively correlated with impact. Some recent studies have revealed steady and universal increase in academic collaborations, however, several important aspects like country-wide variation of collaboration dynamics and effect of collaboration on productivity at global and national level, etc., are not analysed. This paper attempts to bridge this research gap by analysing the global research publication data for the last five decades. The results show a rise in research collaboration globally, though with varying patterns of domestic and international collaboration. The dynamics of collaboration at international level is analysed and the boost in productivity for various countries is computed, indicating varying effects/ dependence of countries in international research collaboration.

Keywords: Collaboration in Science, International Collaboration, Research Collaboration, Scientific Collaboration, Team Science.

1. Introduction

Research collaboration is usually referred to as the working together of researchers to achieve common goal of producing new scientific knowledge (Katz & Martin, 1997). Academia’s interest in research collaboration studies can be traced back to more than six decades when Smith (1958) observed an increase in incidence of multiple-authored papers. Price (1963) used the level/scale of collaboration to distinguish big science and little science. Early identified motivating factors for research collaboration given by Katz & Martin (1997) were (i) very high cost of conducting scientific experiments, (ii) falling of cost of travel and access, (iii) the requirement of social interactions of scientists as one of the organizations of science and (iv) increase in need for specialization (especially for very large research projects that involves specializations in multiple disciplines). In this era the demand for proactive collaboration is in a much better scale and strength, especially for pursuing SDGs (Sustainable Development Goals), the major drivers of collaboration are not limited to mutual benefits in the form of infrastructure sharing and knowledge flows, technology development and transfer, complementary and common solutions for shared problems.

Several studies revealed the notable effects of collaborations on productivity of academia as well as industry. Benefits of academic research collaboration were highlighted by Abramo et al. (2009); Ductor (2015); Parish et al. (2018), etc., while the substantial effect of inventor collaborations on inventor productivity was discussed by Favaro et al. (2012). A spatial institutional analysis (Frenken et al., 2010) for science-based industries revealed that citation impact of research collaboration is higher for international collaboration than for national and regional collaboration. Investigation by Sud & Thelwall (2016) found that international collaboration is not always beneficial while collaboration with some countries like US do guarantee an increase in impact. A recent work by Dong et al. (2017) using scientific publications from 1900 to 2015 revealed the profound shift in research collaborations and citation patterns wherein that the number of collaborative publications had grown 25-fold and number of citations had risen 7-fold during the period, indicating the trend of globalization in scientific collaboration and the citation practice among researchers. Some other studies have also analysed collaboration at international level and have observed that it has rose linearly during the last two 2-3 decades, as measured in terms of number of internationally co-authored papers published (Glanzel, 2001; Persson, Glänzel & Danell, 2004; Lee & Bozeman 2005; Wagner & Leydesdorff, 2005; Leydesdorff & Wagner, 2008; Mattsson et al., 2008; Adams, 2012).

Academic collaborations, be it national or international, happens at different levels, especially at author level and institutional level. Explorations, at different perspectives of author collaborations and its benefits are found in the literature. Major characterization of author collaborations was identified as strong pragmatism and a high degree of self-organization (Melin & Persson, 1996; Wagner & Leydesdorff, 2005). While there are so many studies on research collaboration at specific geographical region over a specific time-period, at the level of specific disciplines or specializations, etc., there are very few studies that analysed long-term dynamics of research collaboration at global level. O'Leary et al. (2008) reported the research collaboration over 50 years in Children’s oncology. Bastos et al. (2021) presented a global bibliometric overview of University-Industry collaboration (UIC) over 50 years wherein the UIC development was divided into four phases and eight research trends were identified. Thelwall & Maflahi (2022) analysed changes in rate of research co-authorship during 1900-2020 for all the 27 Scopus broad fields and 332 Scopus narrow fields. Though their global study revealed continuous and universal increase in academic collaborations, several important aspects like country-wide variation of collaboration dynamics and effect of collaboration on productivity at global and national level, etc., were not analysed. This very important gap is attempted to be addressed in this paper.

Motivated by the above-mentioned gap, we first focus on the analysis of the global collaboration dynamics over the last five decades. This involves analysis of authorship patterns in research publications from the whole world and identification of global collaboration types (domestic vs international collaboration). The collaboration networks at different points of time are visualized and analysed. Apart from this, the dynamics of collaboration network among the top 25 collaborating countries is also very vital, which may help in the identification of strengthening and/or weakening of some ties over time. The effect of collaboration on productivity for top collaborating countries is also investigated in this work. Thus, the objectives of this research can be specified as:

- Investigation of the dynamics of global collaboration during last 50 years,

- Investigation of the dynamics of collaboration network among the top 25 collaborating countries during last 50 years, and

- Investigation of the dynamics of collaborative boost in productivity of top 25 countries during last 30 years.

2. Data and Method

The analysis is based on global research publication data accessed from Dimensions database through a subscription-based access. The period for data was 1970-2020. The major metadata fields that were accessed were those containing year, author details and the affiliating country of the author(s) of a publication. The author collaboration patterns are analysed first by categorizing research publications into different groups: 1-author, 2-9 authors, 10-24 authors and 25+ authors. The publications in these categories over the period were visualised.

Thereafter, the collaborated research output was further grouped into domestic and international collaborations, based on affiliation information of authors (research org country field to be more precise). Next the global research publication data at six points of time (1970, 1980,1990,2000, 2010, and 2020) was used to create international research collaboration network by using VoS viewer (Van Eck & Waltman, 2010). This was done to understand the growth of international research collaboration and identify the major countries participating in international research collaboration at different points of time.

The collaboration patterns for 25 major countries identified in the international research collaboration network was then analysed independently. Research publication data for each country during the 1970-2020 period was accessed for this purpose. The share of non-collaborated, domestic collaboration and international collaboration for all these 25 countries were analysed and plotted.

Finally, a new formalism of productivity boost (Dua, Singh & Lathabai, unpublished) was considered to understand the probable impact of international collaboration on research productivity of the 25 countries. The boost in productivity, indicated by β_p can be defined as:

\(\mathbf{\beta}\mathbf{p}\mathbf{=}\left\lbrack \frac{TP}{TIP} - 1 \right\rbrack\ \times 100\%\) ,

Where, TP stands for the Total Publications of a country (comprising indigenous as well as internationally collaborated papers) while, TIP stands for the Total number of Indigenous Publications of a country. The β_p values for the top 25 countries were computed over six recent intervals of time; 1991-1995, 1996-2000, 2001-2005, 2006-2010, 2011-2015 and 2016-2020. This was done for the recent 30-year period due to the observation of the significant rise of international collaboration during this period.

3. Results

3.1 Authorship patterns at global level

The pattern of authorship in research publications at global level is identified and plotted as area chart in Figure 1. It is observed that the proportion of single-authored papers (non-collaborated) has declined continuously from 1970 to 2020. This decline is from about 50% papers in 1970 to about 30% papers in 2020. On the other hand, the number of collaborated papers show a steady rise during the period. Among the collaborated papers, the highest proportion is that of 2-9 authors, indicating rise in publications from small collaborating teams. The share of papers involving 10-24 authors also shows visible growth since around 1984, though the proportion is low as compared to 2-9 author papers. The share of research papers involving 25+ authors is very small, with minor noticeable presence only after 2010. Thus, it can be said that there has been a continuous rise of collaborated research papers during 1970-2020, with small collaborating teams (2-9 authors) contributing most of the share.

Figure 1: Authorship trends in global research publication data

3.2 Domestic vs International Research Collaboration Trends

The research collaboration can be limited to authors/ institutions within a country or may transcend national boundaries. The figure 2 shows an area-chart of non-collaborated, domestic collaboration and international collaboration during the 1970-2020 period for the global publication data. While the proportion of non-collaborated research output has decreased steadily, the proportion of domestic and international collaboration have both increased (except for a drop in international collaboration proportion during 1986 to 1990).

Figure 2: Proportion of Non-Collaborated, Domestic Collaboration and Internationally Collaborated Papers

3.3 Research Collaboration Networks of top 25 countries at six points of time

In order to further understand the research collaboration at international level over the period of five decades, the research collaboration network graphs at six points of time (1970, 1980, 1990, 2010 and 2020) for the top 25 productive countries are plotted (Figures 3a to 3f). It can be observed that over the period of time, the collaboration networks have become denser. Newer countries are seen to gain prominence in the network for later periods.

Figure 3a: A snapshot of the Research Collaboration network in 1970

Figure 3b: A snapshot of the Research Collaboration network in 1980

Figure 3c: A snapshot of the Research Collaboration network in 1990

Figure 3d: A snapshot of the Research Collaboration network in 2000

Figure 3e: A snapshot of the Research Collaboration network in 2010

Figure 3f: A snapshot of the Research Collaboration network in 2020

For a more systematic analysis of the collaboration networks above, the dynamics of important network related parameters is computed and shown in Table 1. It can be seen that in 1970, there were 204 links and the number of links improved to 252, 284, 299, 300 and 300 at 1980, 1990, 2000, 2010 and 2020, respectively. Average degree improved from 16.32 (at 1970) to 24 (2010 and 2020). Average weighted degree of the network grew from 916.48 (at 1970) to 215132.8 (at 2020), which indicates a rise by almost 235 times. The average path length decreased from 1.32 to 1 signifying the reinforcement of absolute small world effect. Thus, there has been a rise of research collaboration, both in volume and reach, at the international level.

Table 1. Dynamics of collaboration network parameters over six points of time

Parameter/ Year	1970	1980	1990	2000	2010	2020
Number of Links	204	252	284	299	300	300
Average Degree	16.32	20.16	22.72	23.92	24	24
Average Weighted Degree	916.48	2686.24	7935.52	24505.6	76302.64	215132.8
Average Path length	1.32	1.16	1.05	1.003	1	1

3.4 Research Collaboration Patterns of the top 25 countries during 1970-2020

The proportion of non-collaborated, domestic collaboration and international collaboration for the top 25 countries are visualised further to understand the patterns of collaborations of these major countries during 1970-2020 (Figure 4). It can be observed that the proportion of non-collaborated papers in all these countries have decreased with time. However, different patterns are seen for domestic and international collaboration for different countries. While, majority of the countries show a rise in proportion of international collaboration, some countries (China, Japan, Turkey, India) show slightly higher proportion of domestic collaboration. Countries like Germany, France, Netherlands exhibit moderate levels of domestic collaboration and some other European countries like Switzerland, Denmark, Austria, Belgium show higher proportion of international collaboration.

Figure 4: Research Collaboration Patterns of top 25 countries during 1970-2020 (Non-Collaborated, Domestic Collaboration, Internationally Collaborated Papers)

3.5 Productivity Boost of top 25 countries during over 6 intervals of time

In table 2, the β_p values for the top 25 countries are computed and reported over six intervals of time: 1991-1995, 1996-2000, 2001-2005, 2006-2010, 2011-2015 and 2016-2020. The recent period of 30 years is used as it is the period with higher international research collaboration. A higher value of β_p for a country indicates a greater boost in productivity of that country due to collaborations. When β_p > 0%, it is an indication of the existence of collaborations. Moreover, a higher value of β_p implies a significant boost in research productivity of a country as a result of its collaborations. Although, the higher values could also be an indicator of a larger dependence of a country over producing papers through international collaboration, in turn indicating a weak scholarly environment in a country. Thus, an ideal value of β_p is difficult to arrive at. By a rule of thumb, β_p values > 50%, indicate greater reliance of a country on international collaborations for productivity than its own domestic scholarly ecosystem. As β_p values cross 100%, a country is quite dependent on collaborations and needs to make higher efforts to achieve self-sufficiency in research output.

Table 2: Boost in Productivity Values (β_p) for Top 25 countries over six intervals of time

Country	βp values over six intervals of time
	1991-1995	1996-2000	2001-2005	2006-2010	2011-2015	2016-2020
United States	12.44	18.44	24.77	31.24	40.92	52.96
China	31.10	32.18	30.08	30.05	33.17	36.84
United Kingdom	25.73	38.02	55.16	73.42	97.16	148.23
Japan	13.78	19.54	25.77	30.42	36.54	47.78
Germany	34.00	47.00	65.60	78.27	98.41	129.88
France	36.31	50.03	70.81	86.81	110.27	154.33
Canada	28.29	41.51	57.51	69.06	87.82	122.47
Italy	38.00	47.75	56.02	62.31	79.99	103.41
India	11.54	17.32	23.25	24.16	28.70	37.21
Australia	24.26	38.38	54.71	67.51	88.30	132.25
Spain	30.67	41.39	52.22	60.61	80.03	108.33
Russia	18.01	37.52	54.01	48.18	49.47	59.41
Brazil	47.32	46.72	42.18	34.38	42.91	63.25
Netherlands	38.61	58.58	77.65	92.67	126.63	181.06
South Korea	35.58	30.21	32.49	33.53	37.57	44.72
Switzerland	65.82	92.10	122.59	151.13	197.19	270.65
Sweden	47.88	66.11	87.95	113.74	148.98	212.59
Poland	58.73	65.94	61.18	49.74	51.23	65.44
Taiwan	20.63	20.78	23.65	26.01	34.76	61.14
Belgium	56.14	79.72	106.42	126.21	168.50	236.11
Turkey	25.01	22.83	20.68	18.88	23.53	34.56
Iran	48.00	37.18	30.18	24.10	27.57	41.49
Israel	47.76	55.40	62.87	69.98	88.66	107.76
Denmark	54.19	80.53	97.34	118.65	142.44	202.47
Austria	48.27	69.42	93.34	122.22	165.91	224.66

4. Discussion

The analysis of research collaboration shows a global rise in collaborated research output, with varying patterns of domestic and international collaboration for different countries. From the collaboration networks drawn at six points of time, United States is seen to be the most active collaborator among the top 25 productive countries. Starting from the year 1970, United States, United Kingdom and Canada are seen to collaborate actively among each other while United States is also seen to have good collaboration ties with Australia, Germany, Japan and India. Over time in 1980, the prominence of France is observed in the network with collaboration links with countries like Netherlands, Belgium, Italy, Sweden etc. From, 1990 onwards, the collaboration network becomes dense with the emergence of new collaborating partners like Austria, Poland, Switzerland and Denmark. In the years 2000, 2010 and 2020 a more dense and close-knit collaboration network is recorded across all the 25 countries. A dense network of collaboration is found across the European countries while countries such as South Korea, Turkey, Israel etc. slowly mark their presence in the network. Japan, Canada and Spain are seen as examples of other countries rising as prominent players in the collaboration network.

From the results of analysis of dynamics of network, we can clearly identify that the connections (number of collaborative ties) have grown significantly over different time periods until the maximum possible ties is achieved in 2010 (300 (25*12) ties is possible for a network with 25 vertices when it becomes completely connected). Similarly other parameters like average degree, average path length, etc., also converged to maximum possible values at 2010 (when network became completely connected) and remained so in 2020. Average weighted degree kept on increasing unlike the others. This is because the existing ties between all or at least some of the countries further strengthened over the period from 2010 to 2020. Also, if we analyse the dynamics of community structures, in 1970, US, UK, Germany, etc., which had highest collaborative ties were found in one community. Upon the passage of time, when network progressed towards complete connectivity, countries like India and China are found to be in the same community as that of US signalling the growth of collaboration level of both these countries. The observations that (i) China, India, US, Japan, etc., have maintained the boost level below or closer to 50% and (ii) strengthened their international collaborative ties and ended up in same community, is a reflection of their indigenous capability as well as capability to engage with other countries globally.

When it comes to the dynamics of effect of collaborations on productivity of the top 25 countries depicted in table 2, we can see that collaboration always boosted productivity of all the countries though at different degrees. In 1970, India was the country that had the least boost in productivity due to collaboration (11.54 %) and Switzerland benefited most from collaborations then with a boost of 65.82 %. The mean boost for the set of top 25 countries in 1970 was around 35.92. In 1980 also, India and Switzerland maintained their respective positions with boost of 17.32 % and 92.1% respectively while mean boost increased to 46.18%. Switzerland is found to be the country with highest boost in productivity during all the time periods given in table 2. Its latest score is 270.65%. From the period 1980-1990 onwards, Turkey took the place of the country with least boost in productivity from India and remained so till 2010-2020. Average boost due to collaborations in 2020 is found to be 115.16%. Thus, it can be inferred that the degree of effect of collaboration on productivity had also risen like the degree of collaboration for at least top 25 high collaborating countries. It can be seen that most of the European countries in the list of top 25 countries have boost values greater than 100. A productivity boost greater than 50% may signify that a country’s indigenous scholarly ecosystem is not that much contributing towards its productivity or it is more dependent on foreign collaborations for its productivity. Countries like China, India, Japan, US (though in 2020 it has crossed 50% mark), etc., on the other hand seems to have succeeded in strengthening its indigenous research ecosystem possibly through intra-national collaboration. In the long run, the lack of focus on strengthening indigenous research may have some adverse effect on the countries with high boost especially the ones with boost greater than 100. More wholistic picture regarding the effect of collaborations can be drawn only if other boosts such as impact boost, visibility boost, etc., introduced by Dua et al. (unpublished) is computed. To limit the study to the permitted scope, we restrict the analysis to productivity boost only and investigations related to impact and visibility boosts are reserved for future endeavours.

5. Conclusion

The paper presents an analysis of research collaboration at global level during the last 5 decades, looking at both the overall as well as country-specific patterns. The results show a rise in research collaboration globally, though with varying patterns of domestic and international research collaboration. The research collaboration networks drawn are seen to become denser with passage of time, indicating strengthening of existing collaboration ties as well as establishment of new collaboration partners and collaboration ties. The collaboration patterns for the 25 major countries show varying patterns of domestic and international collaboration. Similarly, the boost in productivity due to international collaboration is found to be different for these countries. While countries like China, India, Japan, and US succeeded in maintaining their internal research ecosystem productive while eyeing for the benefits of collaborations on productivity, the dependence of many European countries on international collaboration is noted. While it signals to an extent the high degree of success of these European countries in undertaking collaborative research projects, it may also signify the possible lack of focus on strengthening their indigenous research ecosystem via efforts to foster intra national collaborative ties. More intriguing and vital insights about effects of collaboration can be extricated only if other boosts such as impact boost and visibility boost are computed. Further, the results are persuasive enough to extend this particular analysis over the whole world countries instead of top 25 collaborating countries.

Open science practices

This work is based on analysis of publication data for the world for the period of 51 years (1970-2020) through an API based access of Dimensions database. As such the full publication records were not downloaded, instead the queries were run directly in the database to get various values and intermediate results. The results are prepared and presented using code written in Python language and by using VoS viewer visualization software, both of which are open. We would be happy to share the intermediate results and the python code used for analysis on request.

Author contributions

The first author carried out most of the database fetching and experimental analysis work, in addition to writing a part of the paper draft. The second author conceptualized the study and guided the experimental work and writing. The third author helped in analysis, interpretation and writing of the paper. The fourth author contributed to experimental analysis and visualization of results.

Competing interests

The authors declare that manuscript complies with ethical standards of the conference and there is no conflict of interests whatsoever.

Funding information

This work is partly supported by extramural research grant no.: MTR/2020/000625 from Science and Engineering Research Board (SERB), India, and by HPE Aruba Centre for Research in Information Systems at BHU (No.: M-22-69 of BHU).

References

Abramo, G., D’Angelo, C. A., & Di Costa, F. (2009). Research collaboration and productivity: is there correlation? Higher education, 57(2), 155-171.

Adams, J. (2012). The rise of research networks. Nature 490(7420), 3350336.

Bastos, E. C., Sengik, A. R., & Tello-Gamarra, J. (2021). Fifty years of University-industry collaboration: A global bibliometrics overview. Science and Public Policy, 48(2), 177-199.

de Solla Price, D. J. (1963). Little science, big science--and beyond.

Dong, Y., Ma, H., Shen, Z., & Wang, K. (2017, August). A century of science: Globalization of scientific collaborations, citations, and innovations. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining (pp. 1437-1446).

Dua, J, Singh, V.K., Lathabai, H.H. (unpublished) Measuring and Characterizing International Collaboration Patterns in Indian Scientific Research. Unpublished draft.

Ductor, L. (2015). Does co‐authorship lead to higher academic productivity? Oxford Bulletin of Economics and Statistics, 77(3), 385-407.

Favaro, D., Ninka, E., & Turvani, M. (2012). Productivity in innovation: the role of inventor connections and mobility. Retrieved from: https://mpra.ub.uni-muenchen.de/38950/

Frenken, K., Ponds, R., & Van Oort, F. (2010). The citation impact of research collaboration in science‐based industries: A spatial‐institutional analysis. Papers in regional science, 89(2), 351-271.

Glänzel, W. (2001). National characteristics in international scientific co-authorship relations. Scientometrics, 51(1), pp.69-115.

Katz, J. S., & Martin, B. R. (1997). What is research collaboration? Research policy, 26(1), 1-18.

Lee, S., & Bozeman, B. (2005). The impact of research collaboration on scientific productivity. Social studies of science, 35(5), 673-702.

Leydesdorff, L., & Wagner, C. S. (2008). International collaboration in science and the formation of a core group. Journal of informetrics, 2(4), 317-325.

Mattsson, P., Laget, P., Nilsson, A., & Sundberg, C.J. (2008). Intra-EU vs. estra-EU scientific co-publication patterns in EU. Scientometrics, 75(3): 555-574.

Melin, G., & Persson, O. (1996). Studying research collaboration using co-authorships. Scientometrics, 36(3), 363-377.

O'Leary, M., Krailo, M., Anderson, J. R., & Reaman, G. H. (2008, October). Progress in childhood cancer: 50 years of research collaboration, a report from the Children's Oncology Group. In Seminars in oncology (Vol. 35, No. 5, pp. 484-493). WB Saunders.

Parish, A. J., Boyack, K. W., & Ioannidis, J. P. (2018). Dynamics of co-authorship and productivity across different fields of scientific research. PloS one, 13(1), e0189742.

Persson, O., Glänzel, W. & Danell, R., (2004). Inflationary bibliometric values: The role of scientific collaboration and the need for relative indicators in evaluative studies. Scientometrics, 60(3), pp.421-432.

Smith, M. (1958). The trend toward multiple authorship in psychology. American psychologist, 13(10), 596.

Sud, P., & Thelwall, M. (2016). Not all international collaboration is beneficial: The Mendeley readership and citation impact of biochemical research collaboration. Journal of the Association for Information Science and Technology, 67(8), 1849-1857.

Thelwall, M., & Maflahi, N. (2022). Research coauthorship 1900–2020: Continuous, universal, and ongoing expansion. Quantitative Science Studies, 3(2), 331-344.

Van Eck, N. J., & Waltman, L. (2011). VOSviewer manual. Manual for VOSviewer version, 1(0).

Wagner, C. S., & Leydesdorff, L. (2005). Network structure, self-organization, and the growth of international collaboration in science. Research policy, 34(10), 1608-1618.