s ia, b,1 a,2

Contents

. . . .

. . . .

. . . .A) . .. . . .nalysis. . . .d betwons wit

Tourism Management Perspectives 2-3 (2012) 5564

Contents lists available at SciVerse ScienceDirect

Tourism Managem

j ourna l homepage: www.eAcknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

1. Introduction

Over the past two decades, research collaborations in all disci-plines have grown signicantly, not only at the within-institution

(Glnzel & Schubert, 2005). Scientic collaboration has been shownto benet both research productivity and impact (Eaton, Ward,Kumar, & Reingen, 1999; Inzelt, Schubert, & Schubert, 2009). Themost common form of academic collaboration is co-authorship,level but also at the cross-institutional a

Corresponding author. Tel.: +86 451 86414022; faxE-mail addresses: yeqiang@hit.edu.cn, yeqiang2006@

Haiyan.Song@inet.polyu.edu.hk (H. Song), litonghit@gm1 Tel.: +852 3400 2286.2 Tel.: +86 451 8641 4010.

2211-9736/$ see front matter 2012 Elsevier Ltd. Alldoi:10.1016/j.tmp.2012.03.002tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60roductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614.4. The intensity of research collabora4.5. Network centrality and research p

5. Conclusion . . . . . . . . . . . . . .1. Introduction . . . . . . . . . .2. Literature review . . . . . . .3. Methodology and data . . . . .

3.1. Social network analysis (SN3.2. Data . . . . . . . . . .

4. Cross institutional co-authorship a4.1. Primary data analysis . .4.2. Classic degree centrality an4.3. Linking peripheral instituti. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58eenness centrality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58h mainstream institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59search over the past two deQiang Ye , Haiyan Song , Tong Lia School of Management, Harbin Institute of Technology, 150001, Chinab School of Hotel and Tourism Management, Hong Kong Polytechnic University, Hong Kong

a b s t r a c ta r t i c l e i n f o

Article history:Received 6 September 2011Accepted 28 February 2012

Keywords:Co-authorshipCross-institutional collaborationTourismHospitalityWeighted social networks

This paper examines cross-institutional collaboration in tourism and hospitality research using a co-authorship network model based on papers published in six top-tier tourism and hospitality journals overthe past 20 years. Data analysis reveals that multi-author, multi-university studies are the fastest-growingtype of authorship structure in tourism and hospitality research. We use several network measures to eval-uate the intensity of research collaboration among academic institutions in tourism and hospitality disci-plines. Institutions with tourism and hospitality programs worldwide are ranked according to theircentricity in cross-institutional research collaboration networks and categorized according to their character-istics in collaboration. The empirical analysis shows signicant associations between research performanceand a university's centricity position in cross-institutional research networks. Through advanced social net-work analysis, this study provides new insights into institutional collaboration in tourism and hospitality re-

cades. 2012 Elsevier Ltd. All rights reserved.Reviews in Tourism

Cross-institutional collaboration networknd international levels

: +86 451 86414024.gmail.com (Q. Ye),ail.com (T. Li).

rights reserved.n tourism and hospitality research

ent Perspectives

l sev ie r .com/ locate / tmpwhich is a traditional topic of bibliometric studies.In theelds of tourismandhospitality, bibliometric studies have been

conducted to study the contributions of individual researchers and insti-tutions to the academic community. These include studies of journalrankings (Frechtling, 2004; Jamal, Smith, & Watson, 2008; Park,Phillips, Canter, & Abbott, 2011), assessment of individual research per-formance (Park et al., 2011; Sheldon, 1991), identication of inuential

56 Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564scholars and academic leaders (Zhao & Ritchie, 2007), and evaluation ofresearch themes andmethodologies (Benckendorff, 2009; Palmer, Ses,& Montao, 2005; Rivera & Upchurch, 2008). Some studies have usedthe social network analysis method, such as a study on citation net-works (McKercher, 2008) and several studies on co-authorship net-works (Hu & Racherla, 2008; Racherla & Hu, 2010; Ye, Li, & Law, inpress). These studies provide useful insights into the ways in which re-searchers share, create, and disseminate knowledge in hospitality andtourism research. However, most existing studies in the eld of hospi-tality and tourism management have focused on research collabora-tions at the individual level rather than the institutional level.

With the fast development in scientic research in recent decades,research collaborations in many elds of science, engineering, and so-cial science have crossed university boundaries (Jones et al., 2008). Astudy by Jones and colleagues (2008) published by the elite journalScience showed that multi-university collaborations are the fastest-growing type of authorship structures in most disciplines. This phe-nomenon is attributed, in part, to the development of communicationtechnologies, funding-driven collaboration, changing communicationpatterns, and the increasing mobility of researchers (Glnzel, 2001;Katz & Martin, 1997). In most cases, multi-institutional coauthoredpapers reect the involvement of two or more institutions in re-search. This collaboration will benet participating institutes andaccelerate the diffusion of knowledge. However, empirical studiesindicate that in many elds multi-university collaboration tends toproduce outstanding scientic knowledge in fewer rather than morecenters of high-impact science. A university-level analysis of co-authorship in hospitality and tourism studies will give insight intothe roles, status, and outputs of multi-university collaboration in theeld. It will answer questions like: How do institutions worldwidecollaborate in hospitality and tourism management research? Whatinstitutes are the centers of these collaborative networks? Andhow do multi-university collaborations inuence institutes' researchperformance?

Co-authorship network analysis is now easier than ever. The de-velopment of electronic indexing, archiving of academic publications,and emerging network analysis tools have all made it easier to ana-lyze complicated co-authorship networks. Co-authorship networksare a typical form of social network, formed by authors who havejointly published papers. And co-authorship networks provide impor-tant information on the collaboration among members of the aca-demic community. This study uses advanced social network analysison papers published in six top-tier hospitality and tourism journalsover the past 20 years. The analysis provides new insights into thecollaboration in tourism and hospitality research from an institutionalperspective.

The rest of the paper is organized as follows. The next sectiongives a brief review of published bibliometric studies in tourism andhospitality. Section 3 describes the methodology and data collection.Section 4 presents the research ndings, and the last section summa-rizes the study and provides future directions.

2. Literature review

Bibliometric researchers have been interested in several ques-tions: How is knowledge created and shared within the academiccommunity? What is the best way to evaluate research output?How should the impacts of researchers and institutions be assessed?And how does a particular research eld develop? Bibliometric stud-ies have been conducted to address these questions. Traditional bib-liometric methods include citation analysis and content analysis, butscientic collaboration network analysis has become popular recentlybecause it directly examines cooperation among researchers andresearch institutions. Many studies have contributed to scienticcollaboration research from the co-authorship perspective. Co-

authorship can reliably track many aspects of scientic collaborations(Glnzel & Schubert, 2005) and reects more about the social natureof the academic community than traditional citation analysis.

These analyses have been conducted in a number of elds includ-ing physics, biology, computer science, management, and economics(Acedo, Barroso, Casanueva, & Galn, 2006; Barnett, Ault, &Kaserman, 1988; Newman, 2001a, 2001b). Published studies suggestthat the co-authorship patterns and structures vary among disciplinesand subjects (Newman, 2001a, 2001b, 2001c; Yoshikane, Nozawa, &Tsuji, 2006). Although scientic collaborations can be studied at theindividual, institutional, or national level, most studies have done in-dividual analysis in a specic research area. Fewer studies have inves-tigated institutional collaborations. For example, Matthias and Martin(2004) examined the patterns of co-publications between US universi-ties in economics. They found that geographical distance does not affectmulti-institutional collaborations. In social science research, cross-institutional collaborationswere found to have a citation impact advan-tage over within-institution collaborations, and elite universities domi-nate multi-institution collaborations (Jones et al., 2008).

In the tourism and hospitality eld, bibliometric research hasattracted considerable attention recently. Several studies were con-ducted to investigate research collaboration networks in this specicdiscipline, as well as in the associated academic community (Racherla& Hu, 2010; Tribe, 2010; Xiao & Smith, 2010; Ying & Xiao, in press).Some of the studies were carried out for practical purposes, such astenure qualication and funding distribution (Collison & Sheldon,1991; Pechlaner, Zehrer, Matzler, & Abfalter, 2004; Sheldon &Collison, 1990). Researchers are also interested in identifying aca-demic leaders with a view to understanding knowledge ow in thetourism and hospitality eld (Choi & Sirakaya, 2006; Law & Chon,2007). Of the various types of bibliometric studies, analyses of co-authorship networks in hospitality and tourism have attracted littleattention. Racherla and Hu (Hu & Racherla, 2008; Racherla & Hu,2010) studied co-authorship networks in the hospitality and tourismelds. They mapped individual co-authorship networks, analyzed thedistribution of co-authorships across geographic areas, identied keyresearchers using cohesive groups and structural holes, and calculat-ed correlations between collaboration and publication productivity.Another recent study focused on co-authorship patterns in Australiaand New Zealand tourism research (Benckendorff, 2010). It usedsocial network analysis to examine roles and status of individualresearchers, and it examined cross-institutional and multi-nationalcollaboration in Australian and New Zealand universities. Ye et al.(in press) analyzed individual co-author networks and found that re-searchers' centrality in co-author networks is positively associatedwith their research productivity. Their study also showed that 59.3%of the nodes (authors) in the co-author networks are within themain component.

Although few papers have investigated collaborative networksfrom the institutional perspective, some have evaluated institutionalperformance in hospitality and tourism research using other methods.In the rst such paper, Sheldon (1991) analyzed studies published inthree top tourism journals in the 1980s, identifying leading scholarsand their afliations. Jogaratnam et al. (2005) carried out a similarstudy but extended the journal list and time frame. These two studiesused fourmeasures to evaluate academic institutions: the number of re-search publications, mean productivity per author, portions of outputfrom a particular institution among total output, and number of contri-butions of authors in a given institution. Jogaratnam et al. (2005) in-cluded 11 tourism and hospitality journals in their study, spanning1992 to 2001. Their ndings indicated that institution rankings havechanged considerably over the years. Zhao and Ritchie (2007) rankeduniversities according to the number of leading scholars based on datacollected from eight leading tourism journals from 1985 to 2004. Parket al. (2011) reported the research contribution of universities togetherwith that of individuals and countries in terms of publication productiv-

ity in quality journals (the same journal set used in this study) from

simplicity and robustness. Newman suggested that the collaborationintensity between two authors can be reected by the frequency oftheir collaboration and the total number of authors in each of the col-laborations. The more co-authored papers, the more intense the col-laboration relationship is. In addition, the fewer coauthors on apaper, the more intense the relationship between individual collabo-rators is. Imagine you are a co-author with researcher A on paper Xand a coauthor with researchers B, C, and D on paper Y. You will likelycommunicate more with A than with B, C, and D. Thus it is natural toassume that the collaboration is more intense when fewer collabora-tors are involved. Newman's denition assumes that each collabora-tor makes an equal contribution to the paper and has an equalintensity of collaboration with other coauthors. Despite the unrealisticassumption, the denition reects the combined effect of the two fac-

57Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 55642000 to 2009. They divided the research domain into specic researchstreams and calculated rankings based on a weighted score accordingto the number of authors on a paper instead of instance scores, whichis arguably more objective. Their study shows that institutions contrib-ute differently to specic research streams, and prolic scholars usuallybring prestige to their institutions.

Although bibliometric studies based on counting the number ofpublished articles, as reviewed above, may provide some meaningfulinformation, the insights are still limited by the research methodolo-gies. Emerging network analysis tools and electronic publication datahas made it easier to analyze complicated co-authorship networks.Advanced social network analysis on papers in hospitality and tour-ism journals will provide new insight into collaborations in tourismand hospitality research from an institutional perspective.

3. Methodology and data

3.1. Social network analysis (SNA)

Collaboration is a common social interaction that consists of mul-tiple actors and the relationships between them; this can be repre-sented by the nodes and edges of a network graph. To study acollaborative community in this study, we used the SNA method. So-cial networks are a type of complex network that possess systematicknowledge to describe both static attributes and the dynamic evolu-tion of a particular network structure. Social network analysis hasbeen frequently used in recent years thanks to the development ofnetwork theory and computer processing capability.

In this study, we focus on the position and status of a particularnode in the network structure. SNA uses the term centrality to de-scribe the importance and inuence of a particular actor. Manytypes of centrality metrics have been designed to evaluate the prom-inence of actors from different angles, including degree centrality,betweenness centrality, closeness centrality, reach centrality, eigen-value centrality, ow betweenness, and Bonacich centrality. Mostcentrality measurements stem from the rst three basic measures ofcentrality proposed by Freeman (1978). Degree and betweennesscentralities are the most commonly used measures to evaluate theperformance of actors in a network. Degree centrality is calculatedby summing up the number of edges directly connected to a node,which reects the extent of collaboration of the node. A node on thepath of many links with other nodes is thought to play an importantrole in connecting and transmitting information between differentnodes of the network. Betweenness centrality, therefore, describesthe extent to which a particular node is located on the path of othernodes and is dened as the probability that this node appears onthe shortest path between other nodes. When we take each researchinstitute as a node, the two measurements can be applied to cross-institutional analysis. In this context, each institution becomes a sin-gle knowledge body that can take in and share information withothers through research interactions. In this study, we use thesetwo indicators to identify the roles of institutions in research collabo-ration in hospitality and tourism.

Social networks consist of actors and relationships. Centrality isused to describe actors, and weighted edges are used to describe rela-tionships between the actors. Usually, SNA represents a networkusing an adjacency matrix, with elements represented by 0 or 1 to in-dicate whether a connection exists between two particular nodes.This kind of matrix is used to represent an unweighted network;however, an unweighted network is not useful when one is interestedin the strength of the relationship between nodes. To overcome thisproblem, values are assigned to elements of the adjacency matrixaccording to the intensity of the relationships. The intensity of a rela-tionship can vary. Specically, in a collaborative network, collabora-tion intensity can be calculated in different ways. The Newman

method (Newman, 2001b) is the most popular because of itstors mentioned above that inuence collaboration intensity. In ourstudy, we use Newman's denition to calculate collaboration intensity.

To sum up, SNA is a useful technique that offers a quantitativemeasure of collaboration properties. In particular, including linkweights allows a much deeper understanding of the collaborationamong actors within an academic network.

3.2. Data

To focus our study on the rst-tier, mainstream research commu-nity, we chose the six top tourism and hospitality journals accordingto McKercher, Law, and Lam (2006). The six journals consistentlyranked at the top in similar studies and are frequently used by tour-ism and hospitality bibliometric researchers (listed in Table 1;Frechtling, 2004; Park et al., 2011; Pechlaner et al., 2004; Racherla &Hu, 2010). We retrieved author and afliation information of paperspublished in these journals from 1990 to 2010 using a combinationof web crawling, reference management tools, and manual retrievalfrom ISI, Elsevier, Sage, and EBSCO.

Some databases did not provide information of authors' aflia-tions for articles published in certain years. Therefore, we used onlythe articles for which we could obtain information on author aflia-tions. Table 1 shows the number of papers included in presentstudy and their proportion to the total number of published articlesin the last 20 years for each journal. The dataset consists of 3867 arti-cles, which account for about 65% of all articles published in the sixselected journals during the last 20 years.

We noticed that some databases used different names for thesame institutions. To eliminate these differences, we developed asimilarity comparison program to examine the whole dataset and se-lected all pairs of institutions for which the name had a similarityindex between 0.8 and 1. The afliations of these papers were thenmanually scrutinized to produce a unied list of institution names.In this process, different campuses of the same university were con-sidered to be one institution, except for the University of Wales,which consists of multiple independent institutions. When similarnames appeared, we used information about the country to

Table 1Summary of the dataset used in this study.

Journal Number of articles thatprovided informationon author afliations

Proportion of articlespublished between 1990 and2010 with author afliation

Annals of Tourism Research 884 74%Tourism Management 1292 92%Journal of Travel Research 244 26%Cornell Hotel and RestaurantAdministration Quarterly

279 24%

International Journal ofHospitality Management

748 94%

Journal of Hospitality &Tourism Research

420 100%

Total 3867 65%

determine whether they were the same universities. For example,there are Victoria Universities in Australia, Canada, and New Zealand.Although we made every effort to minimize errors, mistakes may stillremain because of incomplete afliation information.

4. Cross institutional co-authorship analysis

Take The Hong Kong Polytechnic University (HK PolyU) as an exam-ple. A degree centrality of 79 shows that researchers from 79 univer-sities have co-authored papers with HK PolyU researchers. HK PolyU

Fig. 1. Changing co-authorship structures in tourism and hospitality research.

Table 3The top 40 institutions measured by degree centrality and betweenness centrality.

Rank Researchinstitutions

Degreecentrality

Researchinstitutions

Betweennesscentrality(normalized)

1 HK PolyU (China) 79 HK PolyU (China) 19.3752 Nevada 53 Surrey (UK) 8.4723 Virginia Tech 52 Virginia Tech 7.3914 Purdue 51 Central Florida 6.6285 Texas A&M 44 Nevada 6.1886 Pennsylvania State 41 Purdue 5.7587 Surrey (UK) 39 Texas A&M 5.4348 Central Florida 38 Pennsylvania State 5.0439 Cornell 37 Oxford Brookes (UK) 4.79610 Kyung-Hee (Korea) 32 Cornell 4.49311 Michigan State 30 Waterloo (Canada) 3.56612 Oklahoma State 30 College of Charleston 3.52613 Sejong (Korea) 29 Grifth (Australia) 3.43414 Washington State 29 Michigan State 3.21315 Florida State 27 Oklahoma State 3.19316 Temple 27 Calgary 3.17917 Kansas State 26 Florida State 3.10418 Oxford Brookes (UK) 25 Queensland (Australia) 3.02519 College of Charleston 21 Sejong (Korea) 2.844

24 U of Hong Kong(China)

19 Washington State 2.186

25 Calgary 18 Waikato (NewZealand)

2.175

26 Waikato (NewZealand)

18 Catholic U 2.124

27 Iowa State 18 Temple 2.01828 Houston 17 U of Hong Kong

(China)1.923

29 Florida 16 Arizona State 1.92030 Arizona State 15 Western Australia 1.91131 Western Australia

(Australia)15 Leeds (UK) 1.843

32 La Trobe (Australia) 15 Otago (New Zealand 1.77233 Kentucky 14 Newcastle (Australia) 1.74734 James Cook (Australia) 14 Innsbruck (Austria) 1.73035 Guelph (Canada) 14 Florida 1.72836 Ohio State 14 Guelph (Canada) 1.70337 Victoria (Canada) 13 Bournemouth (UK) 1.69938 Strathclyde (UK) 13 Florida International 1.62539 Victoria (Australia) 13 Illinois 1.60240 Colorado State 13 James Cook (Australia) 1.495

58 Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 55644.1. Primary data analysis

Of the 3867 articles in our dataset, 1482 were single-author arti-cles and 2385 were multi-author articles. Of the 2385 multi-authored papers, 1045 papers were co-authored by researchersfrom the same institution, and 1340 were co-authored by researchersfrom different institutions. Thus research collaboration on publica-tions between different research institutions accounted for 56% ofthe total co-authorship output. Multi-institutional coauthored articlesmade up 35% of all publications in the dataset. This indicates thatcross-institutional collaboration is a widespread phenomenon andcontributes signicantly to tourism and hospitality research. Wewill use the 1340 cross-institutional coauthored articles to establishcross-institutional research collaboration networks after Section 4.2(Table 2).

To describe the change in authorship structures over the last20 years, we plotted the different types of author structures (seeFigure 1). Three curves in the gure represent the percentage ofpapers with single author, multi-authors from a single university,and multi-authors from different universities from 1990 to 2010. Athree-year moving average percentage is used to smooth the curves.It clearly shows that the percentage of single-author papers declineddramatically, while multi-institutional co-authored papers increasedsignicantly.

4.2. Classic degree centrality and betweenness centrality

We used the social network analysis program UCINET for thiscross-institutional research collaboration network analysis. Altogeth-er 874 nodes in the cross-institutional collaboration network wereidentied. Each node represents an institute (university) in the hospi-tality and tourism research community. For the initial analysis, wecalculated classic degree centrality and betweenness centrality forall nodes in the entire network. Degree centrality is the total numberof collaborators of a given institution, indicating the range of an insti-tution's collaboration. These two centrality measurements indicatethe importance of each university in the research collaboration net-work. The research institutions were ranked based on these centralityindicators, as shown in Table 3.

Table 3 clearly shows the positions of the top institutions inthe coauthor networks of hospitality and tourism research. Degreecentrality reects the breadth of academic cooperation, whereasbetweenness centrality indicates the core position in knowledge dis-semination of the entire community. Eight universities are at the topof both ranking lists with a little change in order, indicating their im-portance in the knowledge dissemination of hospitality and tourismresearch. Those eight universities are The Hong Kong PolytechnicUniversity, the University of Nevada Las Vegas, Virginia Tech, Purdue,Texas A&M, the University of Surrey, Pennsylvania State University,and the University of Central Florida. Among these top universities,one is in China (Hong Kong), one is in the UK, and six are in the US.

Table 2Primary data analysis.

Author structure All Single-author

Multi-author

Multi-author

Multi-university Single-university

Number of papers 3867 1482 2,385 1340 1045Proportion 100% 38% 62% 56% (35%) 44% (27%)20 Grifth (Australia) 21 Kyung-Hee (Korea) 2.65421 Waterloo (Canada) 20 Wales (UK) 2.55022 Queensland

(Australia)20 Queen Margaret

College (UK)2.364

23 Illinois 19 Strathclyde (UK) 2.201

also has the highest betweenness centrality score, indicating that HKPolyU has the highest possibility of lying on the shortest link of a ran-dom pair of universities. In other words, HK PolyU plays an importantrole in linking different institutions in the worldwide research collab-oration networks of tourism and hospitality studies. HK PolyU is farahead in both rankings, suggesting that researchers there collaboratewidely with researchers from other institutions. In terms of the totalnumber of published papers in the top six journals, these eightuniversities are also ranked at the top of the list (see Table 8).

Universities listed in Table 3 are labeled according to country, ex-cept for universities in the US. Among the top 40 universities on thelist are one university from Austria, two from China (Hong Kong),two from New Zealand, two from South Korea, four from Canada,seven from Australia, seven from the UK, and 25 from the US.

Table 3 shows that some institutions had rather low degree cen-trality but high betweenness centrality. For example, Catholic Univer-sity ranked only 125th in terms of degree centrality, but 26th inbetweenness centrality. The University of Wales (UK) ranked 61stin degree centrality but 21st in betweenness centrality, while the Uni-versity of Leeds (UK) ranked 137th in degree centrality and 31st inbetweenness centrality. Researchers in these universities collaboratewith a relatively small number of universities, but they act as impor-tant bridging nodes for some isolated groups, connecting them to themain academic community. Although these institutions have fewercollaborators (resulting in lower degree centrality), they played animportant role in bridging the gaps in cross-institutional coauthornetworks (resulting in higher betweenness centrality). We use Cath-olic University as an example to visualize its position in the collabora-tive network. Fig. 2 shows that some universities (represented bytriangles) connect to universities in main network (represented bysquares and circles) via Catholic University.

4.3. Linking peripheral institutions with mainstream institutions

Further observation of the networks revealed that many of the 874nodes had only one link to the original network. These scatterednodes are pendants in the network structure that represent peripher-al institutions in worldwide tourism and hospitality research collabo-rations. We simplied the whole network by removing pendants. Werefer to this reduced 508-node network as a pruned network.

Applying component analysis to the pruned network, we obtained11 separate components. The largest component contained 478 nodes,accounting for 94.1% of the total number of nodes. This 478-nodemain component represents the mainstream tourism and hospitalityresearch community.

For each institution, we calculated its degree centrality in the maincomponent of the pruned network and listed them in Table 4 to com-pare it with the degree centrality ranking in the original network(874 nodes) in Section 4.2.

As shown in Table 4, the degree centrality of an institute in themain component is usually smaller than its degree centrality in theoriginal network. The original network represents the entire academ-ic community, whereas the main component of the pruned networkrepresents the mainstream academic community in hospitality andtourism research. The difference in degree centrality measures be-tween the two networks represents the number of peripheral institu-tions that a given institution collaborates with. The magnitude of thatdifference reects its contribution in linking peripheral institutions tothe mainstream academic community. Table 5 lists the institutions forwhich the difference of degree centrality in the original and prunednetworks is larger than three. The ndings suggest that The HongKong Polytechnic University and the University of Central Florida areactive in bridging peripheral institutions with mainstream institutions.

59Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564Fig. 2. Catholic University acts as bridging node in the cross-institutional co-author network.

networks is calculated bywij k ki

kj

nk1, where k refers to the kth co-authored paper, and n refers to the total number of authors in paper k.If i is one of the coauthors of paper k, then ik is assigned a value of 1, oth-erwise 0. It can also be used to examine collaboration between institu-tions, because the relationship between institutions is actually the sumof relationships between individual researchers of those institutions.Using the method proposed by Newman, we calculated the collabora-tion intensity of all pairs of research institutions and transformed theoriginal network into an intensity-weighted collaboration network.Table 6 presents some institutions with high degree centrality in theintensity-weighted network.

We did not use UCINET software to calculate weighted between-ness. To the best of our knowledge, no network analysis tool offers aweighted betweenness centrality calculation, leaving a topic for fur-ther study.

We then calculated the degree centrality of each node in thenewly formed intensity-weighted network and divided it by the de-gree centrality of the node in the original unweighted network. Fora particular node, its unweighted degree centrality is the number ofcollaborators, and its weighted degree centrality is the sum of inten-sity of all its collaborative relationships. The ratio reects the averagecollaboration intensity (ACI) of a particular institution.

ACI Wi Degi weighted

Degi unweighted

Table 4Institutions with the largest degree centrality in the original network (874 nodes) andin the main component of the pruned network (478 nodes).

Rank Top institutions in originalnetwork

Top institutions in the maincomponent

Researchinstitutions

Degreecentrality

Researchinstitutions

Degreecentrality

1 HK PolyU 79 HK PolyU 672 Nevada 53 Virginia Tech. 493 Virginia Tech. 52 Purdue 474 Purdue 51 Nevada 465 Texas A&M 44 Texas A&M 426 Pennsylvania State 41 Pennsylvania State 387 Surrey 39 Surrey 328 Central Florida 38 Cornell 319 Cornell 37 Kyung-Hee 3010 Kyung-Hee 32 Sejong 29

60 Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 55644.4. The intensity of research collaboration

The previous section treated networks as unweighted; that is, allthe edges between different nodes were considered equal. Therefore,when calculating the centrality measures, all the edges were treatedas 1. In this section, we take the intensity of collaboration intoconsideration.

Collaboration is the behavior of knowledge sharing and informa-tion diffusion, so a study of research collaboration networkstosome extentis the study of information ow. When studying collab-orative networks, it is necessary to take into account the magnitude of

11 Michigan State 30 Central Florida 2912 Oklahoma State 30 Oklahoma State 2713 Sejong 29 Florida State 2614 Washington State 29 Temple 2615 Florida State 27 Michigan State 2516 Temple 27 Washington State 2517 Kansas State 26 Kansas State 2318 Oxford Brookes 25 Oxford Brookes 1919 College of Charleston 21 Grifth 1820 Grifth 21 College of Charleston 1821 Waterloo 20 Waikato 1722 Queensland 20 U of Hong Kong 17information ow between nodes. Different levels of afnity betweencollaborators differentiate the ease of information ow betweennodes, which is reected in the collaboration network model bynode-to-node distance. Therefore, collaboration intensity should beadded as a weight in the network.

We dene the intensity of collaboration according to Newman(2001a, 2001b). The intuition behind Newman's method is that col-laboration intensity is positively associated with the number of co-authored papers, but negatively associated with the number ofcoauthors in a particular paper (see the methods section). Specically,Newman's collaboration intensity in cross-institutional research

Table 5Institutions with degree centrality differences greater than 3 between the originalnetwork and the main component of pruned network.

Researchinstitutions

Degree centrality inthe original network

Degree centrality inthe main component

Difference

HK PolyU 79 67 12Central Florida 38 29 9Nevada 53 46 7Surrey 39 32 7Calgary 18 11 7Cornell 37 31 6Oxford Brookes 25 19 6Michigan State 30 25 5Queensland 20 15 5Purdue 51 47 4Washington State 29 25 4Waterloo 20 16 4whereWi is the average of collaboration intensity of node i, andDegi isthe classic degree centrality of node i.

Using this formula, we calculated the ACIs for all 874 nodes anddrew a scatter diagram in which the horizontal axis representsunweighted degree centrality (total number of collaborators), andthe vertical axis represents ACI (Figure 5).

Analysis of the data summarized in Fig. 3 revealed that:

1. ACI values were distributed between 0.25 and 5. However, almostall values were less than 2, except for six institutions with ACIshigher than 2. Among the six exceptions, McGill University wasthe only one to have multiple collaborators (four); the othershad only one collaborator. The reason for their high ACI is that

Table 6The 40 institutions with the highest degree centrality in the weighted collaborationnetwork.

Rank Researchinstitutions

Weighteddegreecentrality

Rank Researchinstitutions

Weighteddegreecentrality

1 HK PolyU 115.500 21 U of Hong Kong 22.0002 Virginia Tech 73.833 22 Florida State 21.6673 Purdue 64.000 23 Oxford Brookes 21.5004 Nevada 58.000 24 Queensland 21.0005 Pennsylvania State 55.600 25 Iowa State 21.0006 Texas A&M 51.333 26 Monash 19.0007 Cornell 51.000 27 Calgary 18.0008 Sejong 46.500 28 Colorado State 17.0009 Surrey 46.333 29 Waikato 17.00010 Central Florida 41.000 30 La Trobe 16.33311 Oklahoma State 39.000 31 Waterloo 16.00012 Washington State 35.333 32 Western Australia 15.33313 Kansas State 33.667 33 Arizona State 15.00014 Temple 32.333 34 Victoria (Australia) 15.00015 Michigan State 31.333 35 South Carolina 14.50016 Kyung-Hee 28.667 36 Ben-Gurion of the

Negev14.000

17 Grifth 26.000 37 James Cook 14.00018 Houston 25.000 38 New South Wales 13.10019 Illinois 24.000 39 Guelph 13.00020 College of

Charleston22.667 40 Hawaii 13.000

research productivity (number of publications) among the 874 insti-tutes in our database. The scores obtained frommethod 1 andmethod2 are both used as a proxy for research productivity.

Table 7

61Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564these researchers collaborated frequently with a limited number ofother researchers, and pairs of individual researchers maintainedintense collaboration. For example, Laurette Dub from McGilland Leo M. Renaghan from Cornell University coauthored fre-quently, resulting in high collaboration intensity between McGillUniversity and Cornell.

2. Among institutions with ACI values less than 0.5, most had onlytwo to six collaborators. Our data showed that these universitiespublished only one or two co-authored papers in the six top jour-nals. Their researchers collaborated infrequently with authorsfrom other institutions, leading to very low ACI values. Most ofthe nodes in this range were peripheral nodes.

3. Almost all institutions considered to be active and productive, asdescribed in the previous sections, had ACI values between 0.8and 1.5. Their unweighted degree centralities were high becausetheir collaboration scopes were much wider. However, they typi-cally had more than two coauthors on each paper; therefore,their ACIs were closer to the average. Although the ACIs of theseinstitutions were generally average, they can still be used as an in-dicator of the closeness of collaboration among institutions. ACIscores also reect the depth and breadth of collaboration. Therankings of institutions with more than 20 collaborators based onACI scores are shown in Table 7 and Fig. 4.

This analysis showed that the institutions can be divided into sev-eral types according to their ACI scores. Type I institutions like SejongUniversity typically have a relatively small number of collaborators,the research productivity of each collaborator tends to be high, but

Fig. 3. Distribution of ACIs of 874 institutions in the collaboration network.the absolute number of output is relatively low. This type of researchinstitution is located on the upper left of the scatter plot (Figure 5).Type II institutions like Florida State University generally have a sim-ilar number of collaborators as Type I. However, most of their collab-orations produce a limited number of papers, leading to low ACIscores. This type of institution is located on the lower left of the scat-ter plot. Type III institutions like HK PolyU and Virginia Tech collabo-rated widely with authors from other institutions, and their ACIscores are also high. These institutions occupy the upper right of theplot. Researchers in type IV institutions like Nevada University alsocollaborate widely with authors from other institutions, but theirACI scores are lower than the type III universities. These institutionsare located in the lower right of the plot.

4.5. Network centrality and research productivity

To explore the relationship between network centrality and aca-demic productivity at the institutional level, we conducted thefollowing analysis. We use the number of published articles torepresent the academic productivity of a university. We evaluatedproductivity using two methods. In the rst method, we simplycount number of papers published by each university. The secondmethod gives partial credit for joint authorship and is used by UTDallas Top 100 Business School Research Rankings.3 Each papercounts for 1 point and is divided equally among all authors on thepaper. Different universities share the points according the numberof authors they have on the paper. For example, if three authors A,B, and C from three universities jointly published a paper, each uni-versity would then receive 1/3 of a point. The top 30 productivityscores are displayed in Table 8. The Pearson correlation coefcientof these two types of productivity counting was as high as 0.992.We then evaluated the relationship between network centrality andproductivity. The Pearson correlation was 0.838 for method 1 and0.788 for method 2, demonstrating a signicant positive correlationbetween network centrality and productivity at the institutionallevel.

We used linear regression analysis to test the relationship be-tween research collaboration (measured by degree centrality) and

Average collaboration intensity ranking of institutions with more than 20 collaborators.

Rank Research institutions ACI Total number ofcollaborators

1 Sejong 1.603 292 HK PolyU 1.462 793 Virginia Tech 1.420 524 Cornell 1.378 375 Pennsylvania State 1.356 416 Oklahoma State 1.300 307 Kansas State 1.295 268 Purdue 1.255 519 Grifth 1.238 2110 Washington State 1.218 2911 Temple 1.198 2712 Surrey 1.188 3913 Texas A&M 1.167 4414 Nevada 1.09 5315 College of Charleston 1.079 2116 Central Florida 1.079 3817 Michigan State 1.044 3018 Kyung-Hee 0.896 3219 Oxford Brookes 0.860 2520 Florida State 0.802481 27Research Productivityi Degree Centrilityi i

Results of the regression model are presented in Table 9, conrm-ing that collaborative activity is highly associated with researchproductivity.

5. Conclusion

With the fast development of academic research and communica-tion technologies, cross-institutional research collaboration has be-come one of the most important types of research collaborations.This paper examines research collaborations among institutions inthe eld of tourism and hospitality studies using the coauthor net-work analysis tools. With multi-authored papers in the six leadinghospitality and tourism journals, this study established cross-institutional networks of co-authorship, which provides insightsinto the roles and status of the research collaboration between

3 http://som.utdallas.edu/top100Ranking/rankingMethod.php

Pennsylvania State University, and the University of Central Florida(among others) are the most productive universities in terms ofpaper publications. Generally speaking, an active university willhave high scores on both degree centrality and betweenness

Type IV

Type IIIType I

Type II

or universities with more than 20 collaborators.

Table 8Institutions with top academic productivity.

62 Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564institutions. The study has accomplished ve tasks that we want tosummarize and highlight.

First, the data showed that multi-university coauthored papers area widespread kind of authorship structure in tourism and hospitalityresearch. Trend analysis indicates that cross-institutional collabora-tion is increasingly becoming the most dominant pattern of researchcollaboration in this eld. The number of multi-university coauthoredpapers is signicantly larger than the number of single-authoredpapers and single-university, multi-authored papers in recent years.This reveals that multi-university coauthored studies have becomeimportant topics to be studied in bibliometric research for tourism

Fig. 4. Average collaboration intensity (ACI) fand hospitality. This primary study simply kicks things off.Second, classic degree centrality and betweenness centrality of

each academic institution are calculated and ranked to assess therole that each institution plays in research collaboration within thetourism and hospitality academic community. Most active universi-ties in worldwide cross-institutional research collaborations are iden-tied and listed. These most active universities include The HongKong Polytechnic University, the University of Nevada Las Vegas,Virginia Tech, Purdue, Texas A&MUniversity, the University of Surrey,

Fig. 5. Relationship between number of collaborators and research performance.Method 1: Simple number counting Method 2: Weighted scores

Researchinstitutions

Number ofpublications

Researchinstitutions

Productivityscore

1 HK PolyU 334 HK PolyU 156.60

2 Cornell 226 Cornell 134.303 Surrey 164 Surrey 82.804 Pennsylvania State 155 Grifth 77.425 Virginia Tech 152 Pennsylvania State 76.086 Grifth 148 Virginia Tech 75.637 Purdue 137 Nevada 67.928 Nevada 134 Central Florida 60.589 Central Florida 115 Purdue 57.5010 Texas A&M 96 Texas A&M 46.5311 Washington State 83 Washington State 39.6712 Kansas State 74 Queensland 38.4813 Michigan State 72 Oxford Brookes 37.8514 Queensland 70 Strathclyde 37.2515 Temple 68 Kansas State 33.0816 Waterloo 66 Waterloo 32.8317 Ben-Gurion of the

Negev62 Michigan State 32.25

18 Oxford Brookes 62 Ben-Gurion of theNegev

30.92

19 Sejong 59 ManchesterMetropolitan

30.23

20 Illinois 58 Hawaii 30.0021 Florida State 57 Temple 29.6722 Strathclyde 56 Illinois 29.5023 Iowa State 54 James Cook 29.5024 Oklahoma State 53 Iowa State 28.5825 U of Hong Kong 51 Shefeld Hallam 27.8326 Hawaii 50 Calgary 27.5327 Manchester

Metropolitan47 Massey 26.58

28 Waikato 47 Florida State 26.5329 Calgary 46 Sejong 24.4830 Shefeld Hallam 45 Waikato 24.20

63Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564centrality. But this study shows that some universities collaboratewith a relatively small group of institutes, while most of their partneruniversities are important and active universities. In this situation theuniversity may have a low degree centrality and a high betweennesscentrality.

Third, we compared metrics of the entire network with the main-stream network. In this way, the study identied some institutionsthat have made signicant contributions in linking peripheral institu-tions to the mainstream academic community. These linking univer-sities include Hong Kong Polytechnic University, the University ofCentral Florida, the University of Nevada, the University of Surrey,the University of Calgary, Cornell University, and Oxford BrookesUniversity.

Fourth, collaborative intensity was added to the originalunweighted network, so that we discovered four apparent types ofcollaborative strategies in terms of breadth and depth of research col-laborations. Type I institutions like Sejong University typically have alimited number of collaborators, but very high research productivityin each collaboration. They usually have high average collaborationintensity (ACI). Type II institutions like Florida State University havea low number of collaborators like Type I universities, but most oftheir collaborations produce a limited numbers of papers, leading tolow ACI scores. Type III institutions like HK PolyU have a largegroup of collaborators and high ACI. They tend to collaborate widelywith authors from other institutions. Meanwhile, most of the collabo-rations result in high productivity. Type IV institutions collaboratewidely with other institutions, but their ACI scores are relative low.

Finally, regression analysis indicates that the centricity of a uni-versity's position in cross-institutional research networks is stronglyassociated with its research performance. On average, one score in-crease in degree centricity will contribute more than three publica-tions to a university. This implies that promoting cross-institutionalresearch collaboration could be a way to improve research perfor-mance for a research institute. Additionally, degree centrality alonecan explain about 90% of the variation of research productivity, mean-ing that a collaboration network perspective does not lose much in-formation compared to traditional counting methods. The socialnetwork perspective keeps the big picture undistorted while stillbeing able to provide many new insights into academic performanceevaluation.

From a macro perspective, exploring cross-institutional collabora-

Table 9Results of linear regression analysis.

Method 1 Method 2

Constant 3.302 (.350)*** 1.540 (.202)***Degree_centrality 3.107 (.049)*** 1.531 (.028)***R2 .908 .880N 874Sig. model .000

Dependent variable: Research Productivity.Number in parentheses is standard error. ***pb .10tion reveals the big picture of academic performanceof the entireacademic community in terms of research collaboration amongthose who research and teach in hospitality and tourism. Understand-ing the interplay among academic institutions sheds new light on theresource allocation models of various institutions. This assessment ofinstitutions in hospitality and tourism research identied inuentialinstitutions and revealed these institutions' contributions to researchcollaboration from a cross-institutional co-author network perspec-tive. Our ndings may also help graduate students in their searchfor employment or research partners from certain types of institutes.Another contribution of our study is that we proposed a new methodof classifying collaboration patterns by investigating how an institu-tion leverages collaboration depth and breadth. It may inspireresearch in other disciplines and thus bring an intensity-weightednetwork perspective to the sight of bibliometric researchers.

Although it adds these insights, this study also suffers from limita-tions. First, a combination of a tourism and hospitality data may un-derplay the status of institutions that work in just one of the twoareas. Second, due to limitations in the available data, two journalsin our dataset have a low rate of author afliation. This may distortthe collaboration network to some extent by neglecting institutionsthat published more frequently in these two journals. With the limita-tion of network analysis software, we did not calculate weighted be-tweenness. These shortcomings can give direction for future studies.

Acknowledgment

This study was partially funded by the Hong Kong PolytechnicUniversity (Grant No.: 1-BB61) and the Fundamental Research Fundsfor the Central Universities (HIT.BRET2.2010013).

References

Acedo, F. J., Barroso, C., Casanueva, C., & Galn, J. L. (2006). Co-authorship in manage-ment and organizational studies: An empirical and network analysis. Journal ofManagement Studies, 43(5), 957983.

Barnett, A. H., Ault, R. W., & Kaserman, D. L. (1988). The rising incidence of co-authorship in economics: Further evidence. The Review of Economics and Statistics,70(3), 539543.

Benckendorff, P. (2009). Themes and trends in Australian and New Zealand tourismresearch: A social network analysis of citations in two leading journals (19942007).Journal of Hospitality and Tourism Management, 16(1), 115.

Benckendorff, P. (2010). Exploring the limits of tourism research collaboration: A socialnetwork analysis of co-authorship patterns in Australian and New Zealand tourismresearch. CAUTHE 2010: Tourism and Hospitality: Challenge the Limits (pp. 151174).

Choi, H. C., & Sirakaya, E. (2006). Sustainability indicators for managing communitytourism. Tourism Management, 27(6), 12741289.

Collison, F. M., & Sheldon, P. J. (1991). An empirical study of employment conditions fortourism and hospitality faculty. Journal of Travel Research, 29(4), 4046.

Eaton, J. P., Ward, J. C., Kumar, A., & Reingen, P. H. (1999). Structural analysis of co-authorrelationships and author productivity in selected outlets for consumer behaviorresearch. Journal of Consumer Psychology, 8(1), 3959.

Frechtling, D. C. (2004). Assessment of tourism/hospitality journals' role in knowledgetransfer: An exploratory study. Journal of Travel Research, 43(2), 100107.

Freeman, L. C. (1978). Centrality in social networks conceptual clarication. Social Networks,1(3), 215239.

Glnzel, W. (2001). National characteristics in international scientic co-authorship re-lations. Scientometrics, 51(1), 69115.

Glnzel, W., & Schubert, A. (2005). Analysing scientic networks through co-authorship. In H. Moed, W. Glnzel, & U. Schmoch (Eds.), Handbook of quantitativescience and technology research (pp. 257276). Netherlands: Springer.

Hu, C., & Racherla, P. (2008). Visual representation of knowledge networks: A socialnetwork analysis of hospitality research domain. International Journal of HospitalityManagement, 27(2), 302312.

Inzelt, A., Schubert, A., & Schubert, M. (2009). Incremental citation impact due to inter-national co-authorship in Hungarian higher education institutions. Scientometrics,78(1), 3743.

Jamal, T., Smith, B., & Watson, E. (2008). Ranking, rating and scoring of tourism journals:Interdisciplinary challenges and innovations. Tourism Management, 29(1), 6678.

Jogaratnam, G., McCleary, K. W., Mena, M. M., & Yoo, J. J. -E. (2005). An analysis of hos-pitality and tourism research: Institutional contributions. Journal of Hospitality &Tourism Research, 29(3), 356371.

Jones, B. F., Wuchty, S., & Uzzi, B. (2008). Multi-University Research Teams: ShiftingImpact, Geography, and Stratication in Science. Science, 322, 12591262.

Katz, J. S., &Martin, B. R. (1997).What is research collaboration? Research Policy, 26(1), 118.Law, R., & Chon, K. (2007). Evaluating research performance in tourism and hospitality:

The perspective of university program heads. TourismManagement, 28(5), 12031211.Matthias, S., & Martin, K. (2004). Patterns of co-authorship among economics departments

in the USA. Applied Economics, 36(4), 327333.McKercher, B. (2008). A citation analysis of tourism scholars. Tourism Management,

29(6), 12261232.McKercher, B., Law, R., & Lam, T. (2006). Rating tourism and hospitality journals. Tourism

Management, 27(6), 12351252.Newman, M. E. J. (2001). Scientic collaboration networks. I. Network construction and

fundamental results. Physical Review E, 64(1), 016131.Newman, M. E. J. (2001). Scientic collaboration networks. II. Shortest paths, weighted

networks, and centrality. Physical Review E, 64(1), 016132.Newman, M. E. J. (2001). The structure of scientic collaboration networks. Proceedings

of the National Academy of Sciences of the United States of America, 98(2), 404409.Palmer, A. L., Ses, A., & Montao, J. J. (2005). Tourism and statistics: Bibliometric study

19982002. Annals of Tourism Research, 32(1), 167178.Park, K., Phillips, W. J., Canter, D. D., & Abbott, J. (2011). Hospitality and tourism re-search rankings by author, university, and country using six major journals: The

rst decade of the new millennium. Journal of Hospitality & Tourism Research,35(3), 381416.

Pechlaner, H., Zehrer, A., Matzler, K., & Abfalter, D. (2004). A ranking of internationaltourism and hospitality journals. Journal of Travel Research, 42(4), 328332.

Racherla, P., & Hu, C. (2010). A social network perspective of tourism research collab-orations. Annals of Tourism Research, 37(4), 10121034.

Rivera, M. A., & Upchurch, R. (2008). The role of research in the hospitality industry: Acontent analysis of the IJHM between 2000 and 2005. International Journal ofHospitality Management, 27(4), 632640.

Sheldon, P. J. (1991). An authorship analysis of tourism research. Annals of TourismResearch, 18(3), 473484.

Sheldon, P. J., & Collison, F. M. (1990). Faculty review criteria in tourism and hospitality.Annals of Tourism Research, 17(4), 556567.

Tribe, J. (2010). Tribes, territories and networks in the tourism academy. Annals ofTourism Research, 37(1), 733.

Xiao, H., & Smith, S. L. J. (2010). Professional communication in an applied tourismresearch community. Tourism Management, 31(3), 402411.

Ye, Q., Li, T., & Law, R. (in press). A co-authorship network analysis of tourism and hospi-tality research collaboration. Journal of Hospitality & Tourism Research. doi:10.1177/1096348011425500.

Ying, T., & Xiao, H. (in press). Knowledge linkage: A social network analysis of tourismdissertation subjects. Journal of Hospitality & Tourism Research. doi:10.1177/1096348011400745.

Yoshikane, F., Nozawa, T., & Tsuji, K. (2006). Comparative analysis of co-authorshipnetworks considering authors' roles in collaboration: Differences between the the-oretical and application areas. Scientometrics, 68(3), 643655.

Zhao, W., & Ritchie, J. R. B. (2007). An investigation of academic leadership in tourismresearch: 19852004. Tourism Management, 28(2), 476490.

64 Q. Ye et al. / Tourism Management Perspectives 2-3 (2012) 5564

Cross-institutional collaboration networks in tourism and hospitality research1. Introduction2. Literature review3. Methodology and data3.1. Social network analysis (SNA)3.2. Data

4. Cross institutional co-authorship analysis4.1. Primary data analysis4.2. Classic degree centrality and betweenness centrality4.3. Linking peripheral institutions with mainstream institutions4.4. The intensity of research collaboration4.5. Network centrality and research productivity

5. ConclusionAcknowledgmentReferences