Investigating Different Types of Research Collaboration and Citation Impact: A Case Study of Harvard University’s Publications1
Ali Gazni Vice president for research affairs of Islamic World Science Citation Center (ISC), Shiraz, Iran; PhD student of Shahid Chamran University, Ahvaz, Iran Address: Jam-e-Jam Ave, Jomhouri Eslami Blvd, Shiraz, Iran Tel: +986273950- Fax: +986270312 E-mail:
[email protected];
[email protected] Fereshteh Didegah PhD. Student in School of Technology, Statistical Cybermetrics Research Group, University of Wolverhampton Address: Wulfruna Street, University of Wolverhampton, Wolverhampotn WV1 1LY, UK E-mail:
[email protected];
[email protected]
Abstract This study aims to investigate the influence of different patterns of collaboration on the citation impact of Harvard University’s publications. Those documents published by researchers affiliated with Harvard University in WoS from 2000 to 2009, constituted the population of the research which was counted for 124,937 records. Based on the results, only 12 percent of Harvard publications were single author publications. Different patterns of collaboration were investigated in different subject fields. In all 22 examined fields, the number of co-authored publications is much higher than single author publications. In fact, more than 60 percent of all publications in each field are multi-author publications. Also, the normalized citation per paper for co-authored publications is higher than that of single author publications in all fields. In addition, the largest number of publications in all 22 fields were also published through inter-institutional collaboration and were as a result of collaboration among domestic researchers and not international ones. In general, the results of the study showed that there was a significant positive correlation between the number of authors and the number of citations in Harvard publications. In addition, publications with more number of institutions have received more number of citations, whereas publications with more number of foreign collaborators were not much highly cited. Keywords: Collaboration patterns, Harvard University, Citation impact.
1
This is a pre-final version of: Gazni, A. & Didegah, F. (2010). Investigating Different Types of ResearchCollaboration and Citation Impact: A Case Study of Harvard University’s Publications. Scientometrics, 87 (2), 251-265.
1
Introduction The complex and interdisciplinary nature of science encourages scholars and scientists to cooperate with one another to gain more advantages through collaboration. According to Patel (1972), scientific collaboration is “a process of functional interdependence between scholars in their attempt to coordinate skills, tools and rewards”. Schrage (1995) also defines collaboration as “the process of shared creation: two or more individuals with complementary skills interacting to create a shared understanding that none had previously possessed or could have come to on their own”. Since 1665 when the first collaborative scientific paper was published, research collaboration was has been extensively in different papers (Luukkonen, Persson & Sivertsen, 1992). According to Wagner-Doebler (2001), at the beginning of the twentieth century, co-authorships accounted for less than 10 percent of all publications, while at the end of the twentieth century, this percentage had gone up to account for over 50 percent of all publications. Abt (2007) notes variations among different sciences in terms of the average number of authors per paper. He states that among 16 examined fields the number of authors per paper ranges from 2.83 authors per paper in mathematics to 9.41 in physics. Different studies have investigated the impact of team work on the citation rate of publications and come to some interesting results (Narin, Stevens and Whitlow, 1991). Generally speaking, research collaboration enhances the quality of research, which leads papers with more authors to be cited more often (Katz and Martin, 1997). Also, a positive correlation between the citation frequency of publications and the number of co-authors of the work has been reported by some researchers (Beaver, 2004; Lawani, 1986; Baldi, 1998). Research collaborations in science are not limited to collaboration between scientists, but also concern collaboration between different organizations, and increasingly, across geographical boundaries among countries (Katz and Martin, 1997), and the results of different research have shown that different types of collaboration can influence research impact (Narin, Stevens and Whitlow, 1991; Schmoch and Schubert, 2008; Sooryamoorthy, 2009). The present study aims to examine the association between the number of citations and the types of collaboration involved in the publications of Harvard University over the period 20002009. Different types of collaboration considered in this study consist of four types including international, domestic, inter and intra-institutional collaboration. Those joint publications, which resulted from cooperation among US researchers and researchers from other countries, were assumed as international publications and those which were just published through collaboration among US researchers were assumed as domestic publications. Publications with inter-institutional collaboration are those resulted from cooperation between Harvard University researchers and researchers from other institutions. Those publications which are co-authored through Harvard University are assumed as publications with intra-institutional collaboration. Research Objectives 2
This study aims to investigate different patterns of research collaboration among Harvard University researchers and their impact on the number of citations received by the publications. In order to achieve the research goal, the following hypotheses are considered: Hypothesis 1. The higher the number of authors per paper, the higher the citation impact. Hypothesis 2. The higher the number of external institutions per paper, the greater the paper’s citation impact. Hypothesis 3. The higher the number of foreign countries per paper, the higher the paper’s citation impact. Literature Review Although collaboration and co-authorship has been studied in many publications, reviewing previous research shows that the study of association between scientific collaboration in a research and the impact of the research results have been investigated in just a small number of studies. As it is not our purpose to give a full review of these studies, only a few have been singled out. Some studies indicate a relationship between an internationally collaborated publication and its citation impact. For instance, Narin, Stevens and Whitlow (1991) found that internationally coauthored papers were cited two times as highly as papers authored by scientists working at a single institution within a single country. Also, Schmoch and Schubert (2008) showed that internationally co-authored publications are highly cited. Glänzel (2001) further pointed out that the world standards indicate a higher expected citation rate of international publications in all fields of science. In another study, Glänzel and de Lange (2002) have analyzed citation patterns of multi-national papers. They concluded that countries generally benefit from participation in multi-national projects. Narin and Whitlow (1990) showed that multiple institution papers are more highly cited than single institution papers, and papers with a foreign collaborator are more highly cited than domestic papers. Katz and Hicks (1997) examined the collaboration strategy that has the largest citation impact. They looked at how citation impact changes in relation to the types of institutions of the authors (in the same institution, other domestic institutions, and foreign institutions), but not the citation differentiation according to disciplines or sectors of the authoring scientists. In one research, Sooryamoorthy (2009) examined the citation patterns of the publications of South African scientists in recent years. The study revealed that the number of citations received by a publication varies not only according to the collaboration, but also to the types of collaboration of the authors who were involved in its production. Based on the results, there was a significant positive correlation between the number of authors and the number of citations. Also, citation rates were significantly higher for publications that involved any kind of collaboration– domestic, intra-institutional, inter-institutional, or international. Research Methodology 3
This study employs the survey research method to answer research questions. ISI Essential Indicators (ESI) is a database which has indexed the most influential institutions, countries, journals and articles in 22 different subject fields. According to this database, Harvard University ranked first among all institutions based on the frequency of citations. We examined all types of Harvard documents published in 9,676 journals introduced by ISI and had a definite broad field (Indeed a collection of 11,327 journals are introduced by ISI2, of which 9,676 journals have a specific ESI subject field). So, all publications included in the mentioned journals, having at least one author with Harvard University affiliation during 1975-2009 have been extracted from ISI Web of Science. During the 35 examined years, Harvard University has published 319,852 documents in the 9,676 journals, of which 196,490 documents were articles, 46,230 documents were meeting abstracts, 17,022 documents were proceedings papers, and 14,606 documents were reviews and other types of documents. Our study was limited to examining Harvard University publications during the last ten years, i.e., 2000-2009. The number of Harvard publications over the mentioned years was 124,937. Different patterns of collaboration are considered as the independent variables, and the normalized number of citations a paper received as the dependent variable. In fact, the impact of the number of authors, number of institutions and number of foreign countries on the normalized number of times a paper was cited were calculated in this research. Different patterns of collaboration were investigated in the 22 different fields of ESI. In citation-based studies, one needs to control the field and the published year (Waltman et al., 2010). Data gathering for the current research was carried out with no consideration of the time window of the citations. In addition, the citation pattern varies significantly across subject fields. Hence, in order to calculate the citation impact of publications, the citation value of each paper was normalized. We have performed a normalization process on the level of papers. To put it more simply, we have divided the number of citations in each paper by the average number of citations of all publications published in the same year and field as the cited paper in the database (this average is offered in the ESI Baseline Table). To calculate the normalized citation per paper (NCPP) index for each group of publications (e.g. single or multi-author publications, single institution publications, two-institution publications etc.), the sum of normalized citations was divided by the total number of publications in the group. Different patterns of research collaboration regarded in the present study include single author, multi-author, international, domestic, inter and intra-institutional collaboration. Linear Regression test and T-Test were used to analyze the data statistically. Research Findings - Distribution of publications and citations in different collaboration patterns To show the noticeable upward trend of scientific collaboration among Harvard University researchers, we investigated the number of authors per paper from 1975 to 2009. As can be seen 2
See: http://sciencewatch.com/about/met/journallist/ to find the list of ISI journals
4
in Figure 1, the average number of authors per paper has increased from 2.58 authors per paper in 1975 to around 10 authors per paper in 2009. Moreover, as Abt (2007) mentions, the average number of authors per paper varies greatly for different sciences. 12
Author Per Paper
10 8
y = 1E-36e0.0424x R² = 0.9568
6 4 2 0 1975
1980
1985
1990
1995
2000
2005
2010
Year
Figure 1. The exponential growth of average number of authors per paper over 1975-2009
Descriptive statistics for publications and citations in each collaboration pattern during the period 2000 to 2009 is given in Table 1. As shown, about 88 percent of publications at Harvard University are multi-author and just above 12 percent are single author publications. The average number of normalized citations received by multi-author publications is about two times more than those by a single author. In other words, single author publications have received one citation on average, while the average number of citations received by multi-author publications is about 2.12 citations per paper. In another section, the collaboration of Harvard University researchers with researchers from other institutions, whether American or non-American, was investigated. The findings revealed that 80.5 percent of publications were the result of author collaboration in different institutions, and about 19.5 percent as a result of intra-institutional collaboration. Calculating the number of citations received by publications with these two collaboration patterns showed that intrainstitutional publications have been cited 1.93 times on average, while inter-institutional publications have received an average number of 2.17 citations per paper. The number of publications co-authored with international and domestic author teams was also calculated. Based on the findings, approximately 69 percent of publications were published through collaboration among national researchers (American researchers), while about 31 percent were internationally co-authored. Investigating the number of citations received by international and domestic publications showed that the average number of citations per paper for domestic publications is 2.03 times per paper, while the international publications were cited 2.32 times per paper, not such a significant difference.
5
Table 1. Frequency and percentage of publications and citation mean in each collaboration pattern
*
Collaboration Pattern
N
% of 124937 (Total Publications)
Single author publications
15075
12.07
Multi-author publications
109862
87.93
Collaboration Pattern
N
% of 109862 (Co-authored Publications)
Intra- institutional collaboration
21422
19.5
Inter-institutional collaboration
88440
80.5
Domestic collaboration
75882
69.07
International collaboration
33980
30.93
NCM* 0.96 2.12 NCM 1.93 2.17 2.03 2.32
NCM stands for Normalized Citation Mean
- Distribution of co-authored and single author publications in 22 subject fields The number and percentage of co-authored and single author publications, sum of normalized citations and normalized citation per paper were investigated in 22 different subject fields. As shown in Table 2, in all 22 fields the number of co-authored publications is much higher than single author publications. In fact, more than 60 percent of all publications in each field (except for Social Sciences, General) are joint publications rather than single author publications. Among all fields, researchers in social sciences have a great tendency to carry out research individually; based on the findings, around 50 percent of publications in social sciences are single author publications. In contrast, researchers in Microbiology have a great tendency toward team work (96.93 percent of their publications are multi-authored). The number of normalized citations per paper (NCPP) in co-authored publications is more than those of single author publications. As shown in Table 2, except for Materials Science, the NCPPs for co-authored publications are higher than those for single author publications. The results of T-test also revealed that there is a significant difference between NCPPs of multiauthor and single author publications in 22 different examined fields (T-value=2.11, df=22, pvalue=0.041; We used an alpha level of .05 for all statistical tests). Table 2. Freq. of co-authored and single author publications and citations in 22 subject fields Single author Publications Field
Multi-author Publications
NCPP**(S)
N
1.08
244
%N of Total N 75.54
NCPP(M)
Agricultural Sciences
79
Biology & Biochemistry
722
7.25
599.79
0.83
9233
92.75
12639.98
1.37
9955
1.65
Chemistry
110
4.04
174.36
1.59
2611
95.96
8355.2
3.2
2721
2.01
Clinical Medicine
4987
9.34
3987.64
0.8
48403
90.66
68108.31
1.41
53390
1.76
Computer Science
128
13.35
128.71
1.01
831
86.65
3330.16
4.01
959
3.99
N
Sum NC* 85.71
Total N
%N of Total N 24.46
6
/NCPP(S)
Sum NC
NCPP(M)
538.77
2.21
323
2.04
Economics & Business
1026
35.35
1186.13
1.16
1876
64.65
4812.61
2.57
2902
2.22
Engineering Environment/Ecology
139
11.7
163.16
1.17
169
13.21
161.18
0.95
1049
88.3
3059.43
2.92
1188
2.48
1110
86.79
1884.32
1.7
1279
1.78
Geosciences
95
6.6
121.46
Immunology
168
4.3
193.56
1.28
1344
93.4
2614.62
1.95
1439
1.52
1.15
3738
95.7
4814.54
1.29
3906
1.12
Materials Science
29
4.68
102.24
3.53
590
95.32
2056.39
3.49
619
0.99
Mathematics
279
Microbiology
59
23.33
350.85
1.26
917
76.67
2659.3
2.9
1196
2.3
3.07
59.93
1.02
1864
96.93
2819.64
1.51
1923
1.49
Molecular Biology & Genetics
450
5.12
399.39
0.89
8345
94.88
12932.33
1.55
8795
1.75
Multidisciplinary Neuroscience & Behavior
458
9.9
2545.41
5.56
4168
90.1
65379.98
15.69
4626
2.82
643
7.78
337.5
0.52
7624
92.22
8949.09
1.17
8267
2.24
Pharmacology & Toxicology
129
10.34
151.51
1.17
1119
89.66
2461.8
2.2
1248
1.88
Physics
324
7.35
566.84
1.75
4085
92.65
11415.3
2.79
4409
1.6
Plant & Animal Science
219
15.59
159.4
0.73
1186
84.41
1572.99
1.33
1405
1.82
Psychiatry/Psychology
1032
16.28
551.46
0.53
5309
83.72
8029.86
1.51
6341
2.83
Social Sciences, general
3795
50.09
2406.74
0.63
3781
49.91
8065.39
2.13
7576
3.36
35
7.45
15.32
0.44
435
92.55
812.09
1.87
470
4.26
Average Value
2.08
Space Science
*
Sum NC stands for Sum of normalized citations. NCPP stands for Normalized citation per paper.
**
- Distribution of publications with inter-institutional and intra-institutional collaboration in 22 subject fields The number and percentage of publications with inter-institutional and intra-institutional collaboration, sum of normalized citations and normalized citation per paper were investigated in different 22 subject fields. Based on the findings, the largest number of publications in all 22 fields were published through collaboration between Harvard researchers and researchers from outside institutions, whether American or non-American. Only in chemistry are approximately half of the publications a result of intra-institutional publications, which shows that Harvard chemists tend to collaborate more with each other at university than in other fields. As can be seen, the number of normalized citations per paper is marginally larger for publications with inter-institutional compared with those of intra-institutional publications (See Table 3). Of course the difference between the two NCPPs is not much significant. The results of T-test also proved this finding (T-value=-0.158, df=42, Sig.=0.87). Table 3. Freq. of publications with inter-institutional and intra-institutional collaboration in 22 subject fields Intra-Institutional Collaboration Field N Agricultural Sciences
28
%N of Total N 11.48
Sum NC* 50.23
Inter-Institutional Collaboration
NCPP**(Ia)
N
1.79
216
7
%N of Total N 88.52
NCPP(Ie)/ Total N
Sum NC
NCPP(Ie)
488.54
2.26
NCPP(Ia) 244
1.26
Biology & Biochemistry
2094
22.68
2379.16
1.14
7139
77.32
10260.82
1.44
9233
1.27
Chemistry
1149
44.01
3321.71
Clinical Medicine
8916
18.42
8120.39
2.89
1462
0.91
39487
55.99
2910.40
81.58
59987.92
1.99
2611
0.69
1.52
48403
1.67
Computer Science
175
21.06
516.46
Economics & Business
304
16.20
634.54
2.95 2.09
656
78.94
2813.70
1572
83.80
4178.07
4.29
831
1.45
2.66
1876
1.27
Engineering
186
17.73
580.57
3.12
863
82.27
2478.85
2.87
1049
0.92
Environment/Ecology
185
Geosciences
250
16.67
279.55
18.60
399.17
1.51
925
1.60
1094
83.33
1604.77
1.73
1110
1.15
81.40
2215.45
2.03
1344
1.27
Immunology Materials Science
662
17.71
137
23.22
843.16
1.27
544.84
3.98
3076
82.29
3971.39
1.29
3738
1.01
453
76.78
1511.55
3.34
590
0.84
Mathematics
90
9.81
170.01
1.89
827
90.19
1658.93
2.01
917
1.06
Microbiology
476
Molecular Biology & Genetics
1779
25.54 21.32
709.06
1.49
1388
74.46
2110.58
1.52
1864
1.02
2331.07
1.31
6566
78.68
10601.26
1.61
8345
1.23
Multidisciplinary Neuroscience & Behavior
879
21.09
1438
18.86
14148.75
16.10
3289
78.91
51231.24
15.58
4168
0.97
1395.48
0.97
6186
81.14
7553.60
1.22
7624
1.26
Pharmacology & Toxicology
189
16.89
289.12
1.53
930
83.11
1073.52
1.15
1119
0.75
Physics Plant & Animal Science
731
17.89
2159.99
2.95
3354
82.11
9255.31
2.76
4085
0.93
311
26.22
316.58
1.02
875
73.78
1256.41
1.44
1186
1.41
Psychiatry/Psychology
789
14.86
1006.79
1.28
4520
85.14
7023.07
1.55
5309
1.22
Social Sciences, general
605
16.00
1107.40
1.83
3176
84.00
6958.00
2.19
3781
1.20
Space Science
49
11.26
74.49
1.52
386
88.74
737.60
1.91
435
1.26
Average Value
1.14
*
Sum NC stands for Sum of normalized citations. NCPP stands for Normalized citation per paper.
**
- Distribution of publications with international and domestic collaboration in 22 subject fields Table 4 shows the frequency of publications with international and domestic collaboration in 22 different fields. Based on the results, in all subject fields except Physics and Space Science, the number of publications with domestic author teams is more than the number of publications with international author teams. Researchers in Social Sciences, Psychiatry/Psychology and Chemistry have a slighter tendency toward international collaboration compared with other sciences. In terms of normalized citations per paper, there is no significant difference between the two patterns. As the results of T-test show, the difference between NCPPs is not significant at 95% confidence level (T-value=0.05, df=42, p-value=0.95). Table 4. Freq. of publications with international and domestic collaboration in 22 subject fields Field
N Agricultural Sciences
140
Biology & Biochemistry
6461
Domestic Collaboration %N of Sum NCPP**(D) Total N NC* 57.38 245.01 1.75 69.98
8634.34
1.34
8
N 104 2772
International Collaboration %N of Sum NC NCPP(I) Total N 42.62 293.76 2.82 30.02
4005.64
1.45
Total N
NCPP(I)/ NCPP(D)
244
1.61
9233
1.08
Chemistry
1951
74.72
4938.49
2.53
660
25.28
1293.61
1.96
2611
0.77
Clinical Medicine
34386
71.04
44665.86
1.30
14017
28.96
23442.46
1.67
48403
1.29
Computer Science
556
66.91
2647.07
4.76
275
33.09
683.08
2.48
831
0.52
Economics & Business
1368
72.92
3687.65
2.70
508
27.08
1124.96
2.21
1876
0.82
Engineering
675
64.35
2060.99
3.05
374
35.65
998.44
2.67
1049
0.87
Environment/Ecology
655
59.01
1068.07
1.63
455
40.99
816.26
1.79
1110
1.10
Geosciences
819
60.94
1445.92
1.77
525
39.06
1168.71
2.23
1344
1.26
Immunology
2463
65.89
3156.52
1.28
1275
34.11
1658.03
1.30
3738
1.01
Materials Science
391
66.27
1563.01
4.00
199
33.73
493.38
2.48
590
0.62
Mathematics
648
70.67
1291.75
1.99
269
29.33
537.18
2.00
917
1.00
Microbiology
1270
68.13
1992.76
1.57
594
31.87
826.88
1.39
1864
0.89
Molecular Biology & Genetics
5511
66.04
7854.12
1.43
2834
33.96
5078.21
1.79
8345
1.26
Multidisciplinary
2818
67.61
42792.87
15.19
1350
32.39
22587.11
16.73
4168
1.10
Neuroscience & Behavior
5243
68.77
6119.59
1.17
2381
31.23
2829.50
1.19
7624
1.02
Pharmacology & Toxicology
699
62.47
881.62
1.26
420
37.53
481.02
1.15
1119
0.91
Physics
1878
45.97
5587.23
2.98
2207
54.03
5828.07
2.64
4085
0.89
Plant & Animal Science
773
65.18
894.08
1.16
413
34.82
678.91
1.64
1186
1.42
Psychiatry/Psychology
3977
74.91
5980.56
1.50
1332
25.09
2049.30
1.54
5309
1.02
Social Sciences, general
2993
79.16
6500.73
2.17
788
20.84
1564.66
1.99
3781
0.91
Space Science
207
47.59
357.73
1.73
228
52.41
454.36
1.99
435
1.15
Average Value
1.02
*
Sum NC stands for Sum of normalized citations. NCPP stands for Normalized citation per paper.
**
- The correlation between number of authors and number of citations Authorship patterns of publications were calculated. Based on the results, Harvard University publications were published with different authorship patterns including one to 2,925 authors per paper. Approximately 87.93 percent of publications were joint publications, of which two-author publications have constituted the largest number of these publications. After that, three-author and four-author publications were the largest number of publications in Harvard University. In order to test Linear Regression on the data, some limited groups were needed. Therefore, those groups of publications with any number of authors which constituted less than one percent of all publications were put into one group. As shown in Table 5, these publications have been collected and placed in the 25th group. So, there are just 25 groups to be analyzed. Based on the findings, the largest number of normalized citation mean belongs to twenty five and more-author publications (Normalized citation mean=7.55) and then twenty two-author publications (Normalized citation mean=7.07). This is while single author publications have been cited less than once on average. All in all, the results of the Regression Test showed that there is a significant correlation between the number of authors in the publications and the frequency of times they were cited (R2=0.9, p