The extent of gender and race/ethnicity imbalance in infectious disease dynamics research

Taube, Juliana 1 ; Merritt, Alexes1; Bansal, Shweta1

Published May 30, 2025 on Dryad. https://doi.org/10.5061/dryad.djh9w0w98

Data files

May 30, 2025 version files 23.18 MB

df9_articledata_0.7_public.csv

8.33 MB
df9_consensus_authorship_public.csv

108.41 KB
df9_consensus_citation_public.csv

14.08 MB
input_jif_regression_nosf_20241212_public.csv

655.24 KB
README.md

6.84 KB

Abstract

Publication practices accumulate to affect credibility and career advancement. Understanding authorship and citation practices is critical to addressing inequities. While citation bias has been demonstrated in several fields, it remains uncharacterized in infectious disease dynamics (IDD), a quantitative, interdisciplinary domain highly visible during the COVID-19 pandemic. We analyze IDD articles and their bibliographies from 2000-2019 using machine-learning algorithms to infer the gender and race/ethnicity of each article's lead and senior author. We examine authorship and citation patterns by gender and racial group across geographic scales, including characterizing the author composition of each article's bibliography relative to the field. Our analysis reveals persistent gender and race imbalances in IDD research. Man-authored and White-authored publications dominate the field, with little progress in racial diversification of US and UK publications over the last two decades. Woman-authored articles have the most representative citation practices but are undercited, especially when women are senior authors. In the US and UK, most citations feature White lead and senior authors, even when citing articles have lead or senior authors of color. These findings underscore the urgent need for more inclusive IDD research practices. We discuss possible mechanisms and solutions to create opportunities for researchers from underrepresented groups.

https://doi.org/10.5061/dryad.djh9w0w98

Description of the data and file structure

This repository provides the source code for the following study: "The extent of gender and race/ethnicity imbalance in infectious disease dynamics research." We also provide anonymized data to reproduce the figures in the text in compliance with our data-sharing agreement with Clarivate.

The data files are already embedded with the appropriate directory structure in the zip file. These data comprise articles in the field of infectious disease dynamics, the inferred gender and race of their authors, the author's country of affiliation, and the Journal Impact Factor where the article was published. We analyze patterns in these data to understand how authorship and citation practices vary across the field.

Data are first pulled from Web of Science using the workflow and files outlined in the pull_data folder. Data are processed using the files in process_data folder; this requires some files published with Dworkin et al.'s paper on citation practices in neuroscience which are copied into this repository. Finally, figures in the main text can be reproduced with the code in make_figures and the data available in the coreidd folder.

The IDD_bib_analysis directory contains the code for our tool to analyze one's bibliography relative to the makeup of the field. Directions are provided in the cleanBib Jupyter Notebook.

Files and variables

Files are labeled numerically according to the order in which they were run within each folder. Across folders the order is 1) pull data, 2) process data, 3) make figures.

Data dictionaries for the data files are provided below.

coreidd/df9_articledata_0.7_public.csv

article_id: a unique identifier for each article
PY: year of publication
PD: month of publication, contains NAs
AG: gender pairing of first and last authors of the article using genderize.io, M stands for man, W for woman, and U for unknown; first letter is first author, second letter is last author
ARbin_ethni: race pairing of first and last authors of the article using ethnicolr and binned into White (W) and non-White (N); first letter is first author, second letter is last author
FA_genderize: first author gender from genderize.io, M stands for man, W for woman, and U for unknown
LA_genderize: last author gender from genderize.io, M stands for man, W for woman, and U for unknown
FA_ethni_race: first author race from ethnicolr, HL stands for Hispanic, W_NL for non-Hispanic White, B_NL for non-Hispanic Black, A for Asian, and U for unknown
LA_ethni_race: last author race from ethnicolr, HL stands for Hispanic, W_NL for non-Hispanic White, B_NL for non-Hispanic Black, A for Asian, and U for unknown
FA_binrace_ethni: first author binary race from ethnicolr, W stands for White, N for non-White (including Hispanic, Asian, and Black non-Hispanic), U is unknown
LA_binrace_ethni: last author binary race from ethnicolr, W stands for White, N for non-White (including Hispanic, Asian, and Black non-Hispanic), U is unknown
Country: country of affiliation for last author of article (string)
global_north: indicator variable for whether the last author's country of affiliation falls in the Global North (0 or 1)
has_single_author: boolean variable for whether the article has only a single author

coreidd/df9_consensus_authorship_public.csv

This file contains the article ids and publication years (PY) for articles in the authorship dataset.

coreidd/df9_consensus_citation_public.csv

This file contains the articles in the citation dataset including which article they were cited by, and whether they should be included in the analysis depending on the threshold of inclusion.

article_id: the article's unique identifier so it can be linked to the full dataset
in_bib_of_article_id: article id of the citing article
PY: publication year of cited article
include_95: boolean for whether article should be included with 95th percentile consensus citation threshold
include_90: boolean for whether article should be included with 90th percentile consensus citation threshold. The main analysis uses this threshold; see the paper for a more detailed description of the construction and interpretation of these thresholds
include_75: boolean for whether article should be included with 75th percentile consensus citation threshold
include_50: boolean for whether article should be included with 50th percentile consensus citation threshold
is_self_cite: indicator for whether the citation is a self citation by the first or last author (0 or 1)

input_jif_regression_nosf_20241212_public.csv,

This file contains data necessary for the impact factor regression in the main text.

pair_gender: gender pairing of first and last authors of the article using genderize.io, M stands for man, W for woman; first letter is first author, second letter is last author
pair_race: race pairing of first and last authors of the article using ethnicolr, W stands for White, N for non-White; first letter is first author, second letter is last author
FA_gender: first author gender from genderize.io, M stands for man, W for woman
LA_gender: last author gender from genderize.io, M stands for man, W for woman
FA_race: first author race from ethincolr, W stands for White, N for non-White
LA_race: last author race from ethincolr, W stands for White, N for non-White
JIF: 2023 Journal Impact Factor of journal where article was published
year: year of publication
num_citations: number of times article was cited by articles in authorship dataset
log_num_citations: natural log of num_citations column
citation_rate: number of citations divided by years since publication
log_citation_rate: natural log of citation_rate column
WC: Web of Science category
num_cited_refs: number of references in the article's bibliography
WC_1st: first entry in WC
Country: country of affiliation for last author of article (string)
global_north: indicator variable for whether the last author's country of affiliation falls in the Global North (0 or 1)
article_id: article unique identifier

Code/software

This code uses both R and Python, depending on the file.

Access information

Other publicly accessible locations of the data:

https://github.com/bansallab/IDD_imbalance

Data was derived from the following sources:

Clarivate Web of Science