Gender differences in authorships are not associated with publication bias in an evolutionary journal

Abstract

The loss of talented women from senior academic positions has partly resulted from a lower number of published papers and the accompanying reduced visibility of female compared to male scientists. The reasons for these gender-differences in authorship is unclear. One potential reason is a bias in the editorial and review process of scientific journals. We investigated whether patterns of authorship and editorial outcome were biased according to gender and geographic location in the Journal of Evolutionary Biology. Such potential bias may contribute to inequality in the field. We found patterns of gender differences in authorship, but this was unrelated to the editorial decision of whether to publish the manuscript. Female first-authors (the lead role) were six times less likely to be named as the corresponding author than male first-authors, and female first-authors were more likely to be displaced as corresponding authors by female co-authors than were male first-authors. We found an under-representation of female first- and last-authors compared to baseline populations of members of the European Society for Evolutionary Biology (which publishes the Journal of Evolutionary Biology) and of Evolutionary Biology faculty at the world top-10 universities for the Life Sciences. Also, manuscripts from Asia were five times more likely to be rejected on the final decision, independent of gender. Overall our results suggest that the peer review processes we investigated at the Journal of Evolutionary Biology are predominately gender-neutral, but not neutral to geographic location. Editorial gender-bias is thus unlikely to be a contributing factor to differences in authorship in this journal.

Fig 1. The percentage of men and women as a) first, b) last and, c) corresponding authors, and broken down by continent of affiliation.

This is based on the 3,348 manuscripts (N: Africa = 46, America = 1051, Asia = 385, Europe = 1664, Oceania = 202) submitted to the JEB between 2012–2016, and the percentages of authors with unassigned gender are also shown.

Fig 2. Factors predicting the gender of the first author, showing the posterior mode estimates for the fixed effects in the binomial model.

The model contained: manuscript type (N: research paper = 1,916, reviews = 88, short notes = 243, special issue = 17; contrast level = research paper), year, continent of affiliation (N: Africa = 26, America = 795, Asia = 82, Europe = 1,208, Oceania = 153; contrast level = Europe), gender of the last author (N: male = 1,697, female = 567; contrast level = female), whether the first author is the corresponding author (N: first author is the corresponding author = 1,818, first author is not the corresponding author = 446; contrast level = first author is not the corresponding author), and the interaction between the gender of the last author and the continent of affiliation. An asterix (*) indicates posterior modes whose 95% credible intervals do not overlap zero, after FDR correction.

Fig 3

Fig 3a. The number of male and female first authors who were or were not the corresponding author. N: female first authors that were corresponding authors = 839, male first authors that were corresponding authors = 1,265, female first authors that were not corresponding authors = 275, and male first authors that were not corresponding authors = 258. The total number of manuscripts does not total 3,348 (the number unique manuscripts assigned a handling editor by the JEB) due to authors with unassigned genders. Fig 3b. The number of male and female first authors who were not the corresponding author and the gender of the corresponding author. N: female first authors displaced by female last author = 77, female first authors displaced by male last author = 157, male first authors displaced by female last author = 49, male first authors displaced by male last author = 163). The total number of manuscripts does not total 3,348 (the number unique manuscripts assigned a handling editor by the JEB) due to authors with unassigned genders.

Fig 4. The number manuscripts with male and female first authors and the gender of the corresponding last author.

N: female first author and female last author = 312, female first author and male last author = 663, male first author and female last author = 255, and male first author and male last author = 1033. Manuscript number does not total 3,348 manuscripts (the number unique manuscripts assigned a handling editor by the JEB) due to authors with unassigned genders.

Fig 5. The percentage of women selected by randomizations from baseline populations of ESEB members and faculty members.

Error bars = 95% confidence intervals. The blue horizontal dashed line represents the percentage of first authors (42%) that were women and the red horizontal dashed line percentage of last authors (25%) that were women on manuscripts sent to the JEB between 2012–2016.

Fig 6. Factors predicting whether the manuscript was sent out for review, showing the posterior mode estimates for the fixed effects in the binomial model.

The model contained: manuscript type (N: research paper = 1,914, reviews = 88, short notes = 242, special issue = 15; contrast level = research paper), year, gender of the last author (N: male = 1,693, female = 566; contrast level = female), continent of affiliation (N: Africa = 26, America = 794, Asia = 81, Europe = 1,205, Oceania = 153; contrast level = Europe), gender of the editor (N: male = 1,920, female = 339; contrast level = female), gender of the first author (N: male = 1,286, female = 973; contrast level = female), and the interaction between the gender of the first author and the editor. * indicates posterior modes whose 95% credible intervals do not overlap zero, after FDR correction.

Fig 7. Factors predicting the first decision (accept/revise or reject) on the manuscript’s outcome, showing the posterior mode estimates for the fixed effects in the binomial model.

The model contained: manuscript type (N: research paper = 1689, reviews = 52, short notes = 200, special issue = 15, contrast level = research paper), year, interaction between the gender of first author and gender of the reviewer, continent of affiliation (N: Africa = 19, America = 683, Asia = 51, Europe = 1066, Oceania = 137, contrast level = Europe), gender of the last author (N: male = 1463, female = 493, contrast level = female), gender of the reviewer (N: male = 1509, female = 447, contrast level = female) and gender of the first author (N: male = 1105, female = 851, contrast level = female). An asterix (*) indicates posterior modes whose 95% credible intervals do not overlap zero, after FDR correction.

Fig 8. Factors predicting the final decision (accept or reject) on the manuscript’s outcome, showing the posterior mode estimates for the fixed effects in the binomial model.

The model contained: manuscript type (N: research paper = 1689, reviews = 52, short notes = 200, special issue = 15, contrast level = research paper), year, interaction between the gender of first author and gender of the reviewer, continent of affiliation (N: Africa = 19, America = 683, Asia = 51, Europe = 1066, Oceania = 137, contrast level = Europe), gender of the last author (N: male = 1463, female = 493, contrast level = female), gender of the reviewer (N: male = 1509, female = 447, contrast level = female) and gender of the first author (N: male = 1105, female = 851, contrast level = female). An asterix (*) indicates posterior modes whose 95% credible intervals do not overlap zero, after FDR correction.