Spatial Network Analysis of Internationalization in Higher Education field
Giovanni Barbato, Aliakbar AkbaritabarThe international co-authorship network of Higher Education (HE) field has been analysed and investigated mainly in descriptive terms, by using a thematic or co-citation analytical approach (Kim et al. 2017). The majority of the studies have investigated this topic by looking at either the internationalization as a research sub-topic of the HE literature (Tight, 2007, 2014; Kosmützky and Krücken; Kuzhabekova et al. 2015) or at the recurrent patterns of collaborations between universities/countries and HE scholars. They have been focused on who occupies the key position and which are the most cited papers and scholars (Jung and Horta 2013). However, this has mostly been analysed only for individual/groups of countries without adopting a global perspective about the co-authorship network and its changes. Furthermore, the growth of the international co-authorship network of HE field have been explained mainly looking at either policy-related and contextual factors such as linguistic commonality and geographic proximity or by the inherent multidisciplinary features of the field (Kim et al. 2017; Avdeed 2019; Wagner and Leydesdorff, 2005). We want to test if and how network mechanisms, related to both structural relationship between authors (e.g. transitivity and preferential attachment) and node level features (differential homophily of research productivity, seniority, affiliation, type of institution, geographical proximity and substantive focus of research), impact on the evolution and growth of the international network of HE field. The empirical analysis is organized in two steps. First, we will look at descriptive statistics of the co-authorship network at the level of authors, organizations (academic institutions [universities and research centres], companies, governments, non-profits, healthcare and other type of institutes), and we will explore their distribution in different country, continent and regional levels providing a spatial bird eye view to temporal changes. From this descriptive analysis, it will be possible to map and underline the international trend of collaboration in HE field at the entire network, organizations and authors level. Second, how network-related mechanisms influence the dynamics of the co-authorship network is investigated by an Exponential Random Graph Model (ERGM). The co-authorship network has been built based on publication data from 54 specialized HE journals retrieved from Scopus in December 2019 and includes 31,431 publications ranging from 1996-2019. Based on data from Scopus, there are 48,076 organizations involved in co-authorship while disambiguation of 24,275 papers reduces those organizations to only 4,122 and number of connected components in the bipartite co-authorship network is decreased from 22,203 to 928. To disambiguate organization names, we use ROR API which uses GRID, ISNI, Crossref and Wikidata. To disambiguate authors, we use Scopus’s Author_ID as an improvement to the raw author information. While being prone to errors (Aman 2018), it offers a somewhat better data quality (a reduction from 66,417 authors to 42,170). Our bipartite community detection setup identifies the most prominent groups of organizations and authors collaborating for a longer time period while our bipartite ERGM uses the node, covariate and structural effects to disentangle the underlying mechanisms that shape the internationalization in Higher Education field.