A study of scientific publishing
Authorshipdisparities
Who submits. Who gets accepted.
American Journal of Public Health · 2025
The study · 2013–2022
Behind every paper,
a submission.
Published articles tell only part of the story. We examined a decade of US manuscript submissions to the American Journal of Public Health, including both accepted and rejected work.
The analysis links each submission to its corresponding author and final editorial decision.
Illustrative manuscripts, not individual records
Names, algorithms and uncertainty
A name is a clue.
Not an identity.
When self-reported demographic information is missing, what can name-based algorithms reveal about publishing disparities?
These are predicted categories, not self-identified identities. The models infer ethnic origin of names and binary gender. Ten imputed datasets carry prediction uncertainty into the analysis.
Predicted race and ethnicity
Who submits.
Who gets accepted.
Submission volume and acceptance tell different stories. Switch between the two measures to see how the pattern changes.
Acceptance was 21.13% for predicted White authors and 14.86% for predicted Asian authors: a 6.27 percentage-point difference.
Name-derived classifications · pyethnicity
Women17.9%acceptance rate
Men20.5%acceptance rate
Gender API · Values rounded to one decimal place
A second dimension
More submissions.
A lower acceptance rate.
Predicted women accounted for more than half of US submissions, but had a lower overall acceptance rate than predicted men.
This descriptive pattern appeared across most predicted racial and ethnic groups. The point estimate for Asian women was slightly above that for Asian men.
Intersecting categories
The widest contrast appears when categories are combined.
Select a group to explore its acceptance rate.
Acceptance rate
White men: 22.99%. Asian men: 14.79%. An 8.20 percentage-point difference between the highest and lowest point estimates.
pyethnicity + Gender API · Point estimates, not a significance ranking
Understanding the pattern
Better data.
A clearer picture.
Algorithms can help journals examine disparities in historical records. Voluntary, self-identified demographic data can make future assessments more informative.
Read the research