Combining topic models with bipartite blockmodelling to uncover the multifaceted nature of social capital

Abstract

In this paper, we seek to identify the existing conceptualisations and applications of social capital contained in the literature, as well as how these are used and combined across and within research fields. Our analytical approach presents a unique combination of topic models and bipartite blockmodelling, enabling us to analyse both the content and structures of a large collection of academic texts. In particular, this allows us to: (a) summarise the content in relation to a variety of topics; and (b) uncover the structure, with diverse text subsets engaging differently with these topics. Our analysis of all of the 11,975 articles on Web of Science that address ‘social capital’ demonstrates that these can be reduced to nine distinct topic clusters and six article clusters. Specifically, we identify the multifaceted nature of the social-capital metaphor and show that there are clear variations in how it is deployed in different bodies of literature. Finally, by mapping the diverse conceptualisations and applications of social capital in a network, we propose a tool for identifying future research opportunities for those interested in novel social-capital treatments in their field.

From colorblind to systemic racism: Emergence of a rhetorical shift in higher education discourse in response to the murder of George Floyd

“…The sum of proportions across all topics totals one for a given institution’s statement. Following Curran et al [ 51 ] and Vlegels and Daenekindt [ 52 ], we dichotomize these edges by coding a tie as having a value of ‘1’ if the proportion is at least twice as high as would be the case if topics were uniformly distributed across statements. As our chosen solution has 18 topics, this means that we coded an edge as 1 if the proportion was at least ((1/18)*2) or 0.1111.…”