Mapping the gender attrition gap in academic psychology

Abstract

Although more women than men enter social science disciplines, they are underrepresented at senior levels. To investigate this leaky pipeline, this study analyzed the career trajectories of 78,216 psychology researchers using large-scale bibliometric data. Despite overall constituting over 60\% of these researchers, women experienced consistently higher attrition rates than men, particularly in the early years following their first publication. Academic performance, particularly first-authored publications, was strongly associated with early-career retention -- more so than collaboration networks or institutional environment. After controlling for gender differences in publication-, collaboration-, and institution-level factors, women remained more likely to leave academia, especially in early-career stages, pointing to persistent barriers that hinder women's academic careers. These findings suggest that in psychology and potentially other social science disciplines, the core challenge lies in retention rather than recruitment, underscoring the need for targeted, early-career interventions to promote long-term gender equity.

Figures (4)

• Figure 1: Gender differences in academic career trajectories. (a) Proportion of women among psychology entrants, by subfield and cohort. (b) Kaplan--Meier survival probability of psychology entrants by gender. (c) Annual transition rates (i.e., the probability of continuing in academia from the previous year) by academic age and cohort group (2000–2004, 2005–2009, 2010–2014). The gray bar indicates years in which retention dropped most sharply. Note that cohort groups differ in their career lengths as a result of the data availability.(d) Summary of gender gaps in transition rates across three time frames: the full captured career span (which varies by cohort), a shared career window common to all cohort groups (academic age 2–9), and the early-career period (academic age 4–6). Positive values indicate higher transition rates for men.
• Figure 2: Statistical description of factors associated with academic performance, collaboration, and institutional affiliations across career stages, disaggregated by gender. Academic performance includes 5-year normalized citation scores and the number of first- and last-authored publications. Collaboration networks are represented by the average impact of co-authors (measured by their 5-year citation scores prior to co-publication), number of co-authors, and proportion of female co-authors. Institutional background is captured by impact of institutional affiliation (proxied by the total 5-year citation scores of affiliated publications), location of the original affiliation, and number of unique affiliations held. Except for the origin region of affiliations (shown as aggregate proportions), all metrics are calculated annually for each researcher by academic age (x-axis) and aggregated to show average trends of each measurement (y-axis) by gender. Note that ‘North America’ here refers to the United States and Canada, excluding Mexico. Shaded areas represent 95% confidence intervals.
• Figure 3: Relative importance of predictors for academic attrition and their temporal dynamics. Blue indicates that a predictor is associated with a decreasing attrition risk over time, while red indicates an increasing attrition risk. Relative importance was calculated by scaling the absolute model coefficients to a 0–1 range, where higher values indicate greater influence on attrition risk. Bar plots (left) show the relative importance of both time-varying and time-variant predictors of attrition among psychology researchers from the cohort groups of 2000–2004, 2005–2009, and 2010–2014. Predictors are ordered by magnitude. Heatmaps (right) show how the relative importance of each time-varying predictor evolved across academic ages. Note that the earliest cohort has the longest career trajectory due to data availability. Predictors are ordered by the academic age at which these variables are most influential: Variables appearing higher tended to shape early-career outcomes, while those appearing lower were more relevant later in researchers’ careers.
• Figure 4: Predicted risk of leaving academia by gender across academic age and cohort group. The predictions average marginal contrasts derived from the time-varying effects model, holding all other covariates constant while varying only gender. Shaded bands indicate the interquartile range (25th to 75th percentile) of the predicted values. Results are shown separately for researchers entering academia in the 2000–2004, 2005–2009, and 2010–2014 cohort groups.