8th International Conference on WIS & 13th COLLNET Meeting, Seoul, Korea Network and productivity: the case of Korean academics 25th October, 2012 Ki-Seok Kwon (KIU), Seok-Ho Kim (NRF) and Duckhee Jang (KSI)
- Contents –1. Introduction2. Literature Review3. Methodology and Data4. Results5. Summary and Conclusion
1. IntroductionAs science becomes bigger, collaborative research is animportant way to accomplish successful scientificresults. More importantly, the creativity of researchresults can be greatly enhanced by the participation ofexperts from various research areas.In a similar vein, a social network as a persuasiveexplanatory variable for individual researcher’sactivities has recently emerged as an importantissue not only in science policy and innovationstudies but also in other social sciences.
2. Literature ReviewThe network of collaborative researcheres have ‘smallworld’ structure (Barabasi et al., 2001). In other words,the distribution of the number of links follows a power lawin most of disciplines.In addition, in spite of the lack of empirical evidences,various characteristics of scientific network tend to beregarded as an influential determinant for scientists’ researchperformance.In order to identify the factors influencing researchproductivity, a number of studies have been carried outduring the last few decades. Scientific productivity isrelated to individual characteristics (e.g. gender, age, anddisciplines) as well as environmental conditions (locations,the amount of funding, and institutional reputations) (e.g.Stephan and Levin, 1992; Stephan, 1996).
2. Literature ReviewHowever, we hardly find a previous study linking these twostreams of studies (i.e. network and productivity).This is probably due to the fact that bibliometric data usuallydo not contain personal information on authors’ gender, ageand disciplines.Against this background, more specific research questionsare as follows.- Is there any disciplinary difference in the network ofcollaborative research network?- Does the network centrality affect the scientists’productivity (and the quality of their papers)?
3. Methodology and DataAfter the investigation of the overall characteristicsof the research network (e.g. the number of linksand node, the number of components, density andmean distance), we implemented an econometricanalysis on whether the network centrality is relatedto the scientific productivityIn order to explore whether the research network of Koreanacademics is a scale-free network or not, we investigate thedistribution of the number of links (i.e. the frequencies of theacademics’ co-authorships). Moreover, various characteristicsof the network are investigated according to discipline. Finally,the network measures such as individual scientists’ centralityare included as an independent variable in a regressionmodel estimating the influence of the position on thescientists’ productivity
3. Methodology and DataWe have collected a very large amount of bibliometric dataof 670,000 published papers in journals monitored andsupported by Korea Research Foundation, which is calledas KCI (Korean Citation Index) data from 2004 to 2009.This data set covers natural science, engineering, socialscience, and humanities fieldsIn this paper, we focus on bibliometric data in the field ofstatistics and computer science (6,614 and 26,510 papersrespectively), as most of papers (i.e. about 80%) in thesetwo disciplines are published in KCI journal rather thaninternational journal (e.g. SCI journal). Moreover, thepapers in the field of economics and public administration(5,254 and 6,000 papers respectively) arecollected, which are mostly published in domesticjournals.
4. Results: Network Characteristics # of # of Average # of Mean Density Nodes Links Degree Comp. DistanceStatistics 1,096 1,208 0.002 1.102 323 9.238Computer 4,781 4,720 0.0002 0.987 1,674 14.099 ScienceEconomics 2110 1,144 0.001 0.542 1,219 11.528 Public 2007 1,055 0.001 0.526 1,210 8.494 Admin.
4. Results: Distribution of # of LinksStatisticians a power’s law Computer scientists
4. Results: Distribution of # of LinksEconomics a power’s lawPublic administration
Network and Research Productivity:statistics and computer science
Network and Research Quality:statistics and computer science
Network and Research Productivity:economics and public administration
Network and Research Quality:economics and public administration
5. Summary and ConclusionFirstly, the collaborative research network of theKorean academics in the field of statistics andcomputer science is scale-free network (i.e. Koreanscholars live in a small world!)Secondly, these research networks show a disciplinarydifference. The network of statisticians is denser thanthat of computer scientists. In addition, computerscientists are located in a fragmented network comparedto statisticians.Thirdly, with regard to the relationship between thenetwork position and scientific productivity, we have founda significant relations and their disciplinary difference.In particular, degree centrality is the strongest predictor forthe scientists’ productivity.
5. Summary and ConclusionBased on these findings, we put forward some policyimplications.Firstly and most importantly, science policy needed to bechanged as the research network is proven to be a ‘small world’.That is to say, scientific resources should be allocated not onlyby the individual academics’ excellence but also by the networkthey are linked to. We may identify a several key journals,organisations and individuals in the network.Secondly, various network indicators measuring the position inthe collaborative research need to be generated and monitored.For example, network centrality indicators can be included inresearch proposals as criteria for funding.Thirdly, we have found some inter-disciplinary areas with aweaker collaborative research, which requires a more intensiveencouragement of the government.
5. Summary and ConclusionIn the future research, we have to expand the scope of thedata set horizontally and longitudinally.In terms of horizontal perspective, as Korean scholars in thefield of natural science and engineering prefer to publish ininternational journals, but the data set based on domesticjournals in these fields do not cover all the papers producedby Korean scholars.In terms of longitudinal perspective, the accumulation of KCIdataset will make time series data available. Moreover, somequalitative data based on interview or document can providericher explanation of the academics’ behaviours in publicationand research collaboration.