Fig. 3: The gender gap in self-promotion is similar across four broad areas but increases with higher performance and academic status quantified by journal impact factor, research productivity, and affiliation rank.

A, The predicted gender gap in self-promotion is of similar magnitude across all four broad research areas. We fit a separate model for each broad area while still controlling for fine-grained subject areas (N = 804,121; 4,123,459; 4,952,083; 2,561,568 for Social, Life, Health, and Physical Sciences; Supplementary Table 10). B–D The likelihood of self-promotion as a function of journal impact factor, author productivity, and affiliation rank, predicted based on our most comprehensive model that also adds an interaction term between gender and the corresponding variable (Supplementary Table 12–14). To quantify research productivity, we decile author’s number of publications as a measure of their productivity (C, a larger bin indicates a more productive decile). Similarly, we decile author’s affiliation rank (D, a larger bin indicates a higher rank decile). The three regressions in B–D are fitted to all 11,396,752 (paper, author) pairs. Predictions are based on setting all control variables at their median values in our data. Error bars indicate 95% bootstrapped confidence intervals.