Fig. 2 Learned structure of the Bayesian network. For the sake of readability, we did not report the scientific field effect, which was linked to all nodes. Orange arrows indicate a negative relationship, and blue arrows indicate a positive relationship (dotted black, if the sign depends on the scientific field taken into consideration). Path coefficients are only shown for paths that were consistent across scientific fields. All path coefficients can be found in table S9.
- Table 1 Number of journals and frequency distribution of selected sample characteristics by field of research.
Biomedicine and health Life sciences Physical sciences Social sciences and
humanitiesNumber of journals 55 24 50 16 Mean impact factor (SD) 2.99 (1.49) 3.14 (1.60) 3.04 (1.32) 2.18 (1.07) Number of submissions 113,421 31,331 184,315 19,051 Percentage first-round
rejections45.8 35.2 41.2 50.0 Percentage final rejections 58.8 48.1 48.5 62.3 Percentage women authors 31.5 27.7 19.1 38.0 Percentage women referees 24.6 21.0 16.3 38.1 - Table 2 Logistic mixed-effects models on the final editorial decision (accept) by field of research using the gender ratio as predictor.
Mean estimate, 95% CI, and Bayes factor (β > 0) are reported for each variable.
Variable Biomedicine and health
scienceLife science Physical science Social science (Intercept) −6.224 −4.698 −7.069 −5.124 [−6.629, −5.827] [−6.048, −3.366] [−7.970, −6.174] [−6.071, −4.200] 1:20,000 1:20,000 1:20,000 1:20,000 Women proportion (authors) 0.129 0.050 0.205 −0.065 [0.022, 0.235] [−0.143, 0.244] [0.115, 0.296] [−0.291, 0.156] 103:1 2:1 20,000:1 1:2 Women proportion (referees) −0.154 −0.042 −0.041 −0.234 [−0.240, −0.070] [−0.206, 0.122] [−0.119, 0.036] [−0.448, −0.020] 1:2,856 1:2 1:6 1:59 Review score 6.020 6.176 6.095 5.823 [5.907, 6.134] [5.936, 6.416] [5.996, 6.194] [5.470, 6.181] 20,000:1 20,000:1 20,000:1 20,000:1 Agreement 1.214 0.667 0.708 0.202 [1.086, 1.339] [0.449, 0.879] [0.613, 0.801] [−0.122, 0.525] 20,000:1 20,000:1 20,000:1 8:1 IF −0.059 −0.140 0.058 −0.143 [−0.112, −0.004] [−0.215, −0.065] [0.020, 0.095] [−0.403, 0.114] 1:57 1:20,000 832:1 1:6 Number of authors 0.002 −0.039 0.045 0.014 [−0.006, 0.011] [−0.053, −0.025] [0.035, 0.054] [−0.026, 0.055] 2:1 1:20,000 20,000:1 3:1 Number of referees −0.184 −0.160 −0.103 −0.300 [−0.226, −0.142] [−0.234, −0.0986] [−0.133, −0.072] [−0.420, −0.180] 1:20,000 1:19,999 1:20,000 1:20,000 PR type: single-blind 0.532 0.117 1.185 1.091 [0.97, 0.962] [−1.228, 1.472] [0.281, 2.110] [−0.391, 2.592] 105:1 1:1 162:1 14:1 Number of revision rounds 4.094 3.670 3.99 3.756 [4.037, 4.152] [3.578, 3.766] [3.95, 4.04] [3.624, 3.889] 20,000:1 20,000:1 20,000:1 20,000:1 Sensitivity 0.93 0.93 0.93 0.92 Specificity 0.96 0.95 0.97 0.97 - Table 3 Logistic mixed-effects models on the final editorial decision (accept) by field of research using the first and last author’s gender as predictors.
Mean estimate, 95% CI, and Bayes factor (β > 0) are reported for each variable.
Variable Biomedicine and health
scienceLife science Physical science Social science (Intercept) −6.116 −4.502 −7.020 −5.291 [−6.530, −5.700] [−5.844, −3.156] [−7.960, −6.088] [−6.282, −4.322] 1:20,000 1:20,000 1:20,000 1:20,000 First author woman 0.001 −0.099 0.099 −0.065 [−0.067, 0.069] [−0.218, 0.022] [0.035, 0.163] [−0.259, 0.127] 1:1 1:18 768:1 1:3 Last author woman −0.056 −0.050 −0.034 0.039 [−0.125, 0.014] [−0.181, 0.081] [−0.109, 0.024] [−0.148, 0.223] 1:16 1:3 1:8 2:1 Women proportion (referees) −0.135 −0.063 −0.033 −0.190 [−0.233, −0.037] [−0.254, 0.130] [−0.132, 0.066] [−0.429, 0.044] 1:302 1:3 1:3 1:16 Review score 6.017 6.246 6.056 5.785 [5.889, 6.145] [5.966, 6.532] [5.928, 6.186] [5.393, 6.181] 20,000:1 20,000:1 20,000:1 20,000:1 Agreement 1.207 0.635 0.646 0.353 [1.063, 1.353] [0.387, 0.886] [0.523, 0.769] [−0.003, 0.710] 20,000:1 20,000:1 20,000:1 38:1 IF −0.059 −0.139 0.044 −0.173 [−0.120, 0.002] [−0.223, −0.053] [−0.003, 0.091] [−0.454, 0.113] 1:33 1:1,817 28:1 1:8 Number of authors 0.005 −0.045 0.051 0.024 [−0.005, 0.015] [−0.061, −0.029] [0.039, 0.063] [−0.020, 0.068] 6:1 1:20,000 20,000:1 6:1 Number of referees −0.188 −0.199 −0.137 −0.286 [−0.236, −0.141] [−0.285, −0.114] [−0.177, −0.098] [−0.416, −0.155] 1:20,000 1:20,000 1:20,000 1:20,000 PR type: single-blind 0.537 0.099 1.336 1.094 [0.113, 0.974] [−1.245, 1.435] [0.405, 2.284] [−0.406, 2.601] 143:1 1:1 391:1 14:1 Number of revision rounds 4.100 3.707 4.018 3.834 [4.036, 4.165] [3.597, 3.819] [3.961, 4.076] [3.687, 3.988] 20,000:1 20,000:1 20,000:1 20,000:1 Sensitivity 0.93 0.93 0.93 0.92 Specificity 0.97 0.96 0.97 0.96 - Table 4 Poisson regression model predicting the number of rounds of reviews before manuscript’s acceptance.
Mean estimate, 95% CI, and Bayes factor (β > 0) are reported for each variable.
Variable Biomedicine and health
scienceLife science Physical science Social science (Intercept) 0.571 −1.171 −1.537 −1.427 [0.488, 0.654] [−1.444, −0.894] [−1.821, −1.255] [−1.710, −1.149] 20,000:1 1:20,000 1:20,000 1:20,000 Women proportion (authors) −0.002 −0.001 0.016 −0.006 [−0.031, 0.027] [−0.072, 0.071] [−0.019, 0.052] [−0.080, 0.069] 1:1 1:1 4:1 1:1 Women proportion (referees) 0.037 0.083 0.049 0.007 [0.013, 0.060] [0.022, 0.143] [0.019, 0.079] [−0.069, 0.082] 951:1 307:1 951:1 1:1 Review score −0.389 1.712 1.812 2.251 [−0.423, −0.355] [1.642, 1.783] [1.783, 1.842] [2.153, 2.349] 1:20,000 20,000:1 20,000:1 20,000:1 IF −0.007 0.021 0.036 0.026 [−0.021, 0.007] [−0.002, 0.043] [0.022, 0.049] [−0.06, 0.111] 1:6 28:1 20,000:1 3:1 Number of authors 0.002 0.010 0.014 −0.005 [0, 0.005] [0.004, 0.015] [0.01, 0.017] [−0.018, 0.008] 43:1 6,666:1 20,000:1 1:3 Number of referees 0.053 0.065 0.106 0.089 [0.042, 0.063] [0.039, 0.091] [0.095, 0.117] [0.051, 0.127] 20,000:1 20,000:1 20,000:1 20,000:1 Agreement −0.022 0.031 0.056 0.033 [−0.055, 0.012] [−0.048, 0.111] [0.019, 0.092] [−0.075, 0.145] 1:9 4:1 799:1 3:1 PR type: single-blind −0.072 −0.051 0.095 −0.133 [−0.155, 0.011] [−0.303, 0.196] [−0.192, 0.383] [−0.554, 0.278] 1:22 1:2 3:1 1:3
Supplementary Materials
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/2/eabd0299/DC1
Additional Files
Supplementary Materials
Peer review and gender bias: A study on 145 scholarly journals
Flaminio Squazzoni, Giangiacomo Bravo, Mike Farjam, Ana Marusic, Bahar Mehmani, Michael Willis, Aliaksandr Birukou, Pierpaolo Dondio, Francisco Grimaldo
This PDF file includes:
- Data sharing protocol
- Figs. S1 to S3
- Tables S1 to S11
Files in this Data Supplement:
