This document discusses China's scientific performance based on bibliometric analysis. It finds: - China's scientific output and impact has grown tremendously, with its share of world publications rising from 3% in 2000 to 17% in 2015. - Chinese research is particularly strong in physical sciences, engineering, mathematics and computer science. - Analysis of individual institutions like Zhejiang University and Fudan University reveals their research strengths in specific micro-level research areas. - The document advocates for responsible use of bibliometrics and more detailed analyses to provide context beyond high-level statistics.

1. An in-depth bibliometric perspective on China’s
scientific performance
Ludo Waltman
Centre for Science and Technology Studies (CWTS), Leiden University
Fudan University
April 12, 2019

2. Centre for Science and Technology Studies (CWTS)
• Research center at Leiden University in the Netherlands
• Science and technology studies, with a strong emphasis
on scientometrics
• About 50 staff members
• Our mission: Making the science system better!
1

3. Outline
• Country-level analysis
• Institutional-level analysis
• Visualization of science using VOSviewer
• Algorithmic identification of micro-level research areas
• Taking a closer look at China’s scientific performance
• Responsible use of scientometrics
2

4. Country-level analysis
3

5. Scientific output
International comparison over time – absolute perspective
4

6. Scientific output
International comparison over time – relative perspective
• Huge growth in China’s
share in world output, from
3% in 2000 to 17% in 2015
• This causes a decline in
other countries’ share in
world output
5

7. Scientific output
International comparison between fields (publications from 2015)
• China has a strong focus on
physical sciences and
engineering as well as
mathematics and computer
science
• China makes a relatively
small contribution to the
international literature in the
social sciences
6
Normalized output above (below) 1.0 means that
output is larger (smaller) than expected based on
the size of a country and a field

8. Scientific impact
International comparison over time
• Large increase in China’s
percentage of highly cited
publications, from 5% in
2000 to 8.5% in 2015
7A publication is highly cited if it belongs to
the top 10% most cited of its field

9. Scientific impact
International comparison between fields (publications from 2015)
• Scientific impact of Chinese
publications in mathematics
and computer science,
physical sciences and
engineering, and social
sciences is close to the world
level
• Chinese publications in
biomedical and health
sciences have a low
scientific impact
8
A publication is highly cited if it belongs to
the top 10% most cited of its field

10. Institutional-level
analysis
9

11. CWTS Leiden Ranking
• Bibliometric ranking of major
universities worldwide
• Advanced methodology based
on most recent insights from
bibliometric research
• www.leidenranking.com
10
Data source: Clarivate Analytics Web of Science

12. Chinese universities
11
Fuzhou Univ.
Hong Kong Univ.
Science and Tech.
City Univ.
Hong Kong
Zhejiang Univ.
Shanghai Jiao
Tong Univ.
Tsinghua Univ.
Peking Univ.
Fudan Univ.

13. Chinese universities vs. US universities
12
Rockefeller Univ.
Harvard Univ.
Johns Hopkins
Univ.
Univ. Michigan
Stanford Univ.MIT

14. Chinese universities vs. Dutch universities
13
Leiden Univ.

15. Visualization of
science using
VOSviewer
14

16. VOSviewer
15

17. Development of VOSviewer
16

18. Data sources supported by VOSviewer
17

19. Users of VOSviewer
• Researchers
• Research institutions
• Research funders
• Scientific publishers
• Industry
18

20. Algorithmic
identification of micro-
level research areas
19

21. Community detection in networks
20

22. Louvain algorithm
21

23. Louvain algorithm: Overview
22

24. Louvain algorithm: Disconnected communities
23

25. Leiden algorithm
24

26. Leiden algorithm: Overview
25

27. Leiden algorithm: Fast local moving
26
Instead of iterating over all nodes, fast local moving iterates
only over unstable nodes; node i is considered unstable if at
least one of its neighbors has moved after the most recent
move of node i itself (Ozaki et al., 2016)

28. Properties of the Louvain and Leiden algorithms
27
Louvain Leiden
Every iteration
𝛾-separation + +
𝛾-connectivity - +
Stable iteration
Node optimality + +
Subpartition 𝛾-density - +
Asymptotic Subset optimality - +
Subset optimality: There is no subset of nodes in any community
that can be moved to a different (or a new) community

29. Comparison of the Louvain and Leiden algorithms
28

30. Identifying micro-level research areas
29
• Data:
– CWTS in-house version of Clarivate Analytics Web of Science database
– Time period: 2000–2017
• Method:
– Construct citation network of publications
– Use Leiden algorithm to identify communities (i.e., research areas) in this network
– Label research areas based on terms extracted from titles of publications

31. Micro-level research areas
30
Social sciences
Biomed.
& health sciences
Math. & comp.
science
Physical
sciences & eng.
Life & earth sciences
• Subdivision of science
into 4000 research
areas

32. Taking a closer look at
China’s scientific
performance
31

33. China’s research strengths
• China is particularly
strong (i.e.,
responsible for
≥ 10% of all highly
cited publications) in
green research areas
• China’s research
strengths can be
found mostly in
physical sc. and eng.
and math. and comp.
sc.
32
Social sciences
Biomed.
& health sciences
Math. & comp.
science
Physical
sciences & eng.
Life & earth sciences

34. 33
Germany
China
USAUK

35. Zhejiang University’s research strengths
34
Share of an area’s
highly cited
publications
authored by Zhejiang
University

36. Zhejiang University’s research strengths
35
Share of an area’s
highly cited
publications
authored by
Zhejiang University
bibliometrics
China

37. Research area 1127: Topics
36

38. Research area 1127: Focus of Zhejiang University
37

39. Research area 145: Topics
38

40. Research area 145: Focus of Zhejiang University
39

41. Fudan University’s research strengths
40
Share of an area’s
highly cited
publications
authored by Fudan
University
microRNA

42. Research area 2: Topics
41

43. Research area 2: Temporal development
42

44. Research area 2: Focus of Fudan University
43

45. Responsible use of
scientometrics
44

46. Diana Hicks,
Paul Wouters,
Ludo Waltman,
Sarah de Rijcke,
Ismael Rafols
Nature,
April 23, 2015,
520, 429–431

47. 46
www.leidenmanifesto.org

48. Korean
German
Persian
Catalan
Finnish
Translations into many languages
Chinese
Spanish
French
Brazilian
Portuguese
Japanese
Traditional
Chinese
Russian
Basque
Italian

49. 48

50. Conclusions
49

51. Conclusions
• Impressive growth in China’s scientific performance, not only in terms of
output but also in terms of impact
• China has a strong focus on physical sciences and engineering as well as
mathematics and computer science
• Deeper understanding of China’s scientific performance requires more
detailed bibliometric analyses, for instance focused on research strengths of
individual institutions
• Bibliometrics is often too much focused on high-level statistics; truly
interesting insights are obtained from more detailed analyses that provide
context to high-level statistics
• Use bibliometrics responsibly!
50

52. Thank you for your attention!
51