Social Capital Visualization

Using bibliometric analyses to visualize fractures in social capital research. Paper published in Socius w/ Kate Vinita Fitch. The basic idea here is to provide a preliminary characterization of the existing divides and lack of consensus around conceptualization and usage of social capital as a concept.

This repo provides the code (but not the data)¹ for replicating the analyses in the manuscript.

Basic Steps -

0. As with most of these types of analyses in the work I've done w/ Ryan Light (see details here), I start from ISI. There's a screengrab version of my search criteria here.² The search³ narrowly defined the corpus as all documents including "social capital" as a topic.⁴
1. 1_DataProcessing.R - I used the bibliometrix package to process those files into dataframe format.
2. 2_BuildNetworks.Rmd - I used an old script of mine Scientometric_Networks.R (in the scripts folder) to construct citation, co-citation, and bibliographic coupling networks from these data.
   • Dependencies: social_capital.Rdata (created in step 1).
   • As an intermediate step here, there is some manual data cleaning of the "key citations" to ensure each citation of them is combined, using this file.
3. 3_NetworkCommunities.R - Using basic functions in the igraph package, I identify community solutions for the citation network, then extract co-citation rates within those communities, and compare those to base-rate expectations.
   • Dependencies: networks.Rdata (created in step 2).
4. 4_TablesFigures.Rmd - This file draws on the elements compiled from the steps above to build the analyses actually presented in the paper, and some supplementary information.
   • Dependencies: many of the objects created in the various steps above (including those in the bibmx_networks.Rdata, networks.Rdata, and clusters.Rdata environments as well as the louvain.rda object).

Included Data Objects:

1. data/louvain.Rda - the primary results from the louvain solution that is presented in the paper. This environment includes 3 objects:

   • lc_cites - the number of citations received by each "key paper" from papers in each of the identified communities (this provides the information for the top panel of Figure 1).
   • lcct - the frequency of co-citations among the "key papers" within each of the identified communities.
   • lc_z - the computed z-scores for the divergence from random expectations (see #3 below) of those co-citation rates (this provides the information for the bottom panel of Figure 1).

2. data/fast_greedy.Rda - the corresponding elements to #1, but using the fast greedy community solution; here the "lc" prefixes above are all replaced with "fg"s.

3. baseline.Rdata - this includes a single object that is a list of 1000 draws of random co-citation rates among the "key papers" in the corpus. See a text explanation of how this baseline was determined in the Appendix, or the code for computing it in the 3_NetworkCommunities.R script.

Notes

Footnotes

1. To comply with Web of Science data use agreements, this repo does not include raw, nor compiled datasets of complete records. ↩

2. If replicating, be sure to grab the "Full Record and Cited References" versions of these files. ↩

3. This was executed on Oct 23, 2020, and is constrained on the front end by my university's subscription to ISI (only goes back through 1974). But I can't imagine either of those really determine what we find here. ↩

4. The ISI "topic" field includes searching the title, abstract, and keywords of each document. ↩