Despite increasing representation in graduate training programs, a disproportionate number of women leave academic research without obtaining an independent position that enables them to train the next generation of academic researchers. To understand factors underlying this trend, we analyzed formal PhD and postdoctoral mentoring relationships in the life sciences during the years 2000 to 2020. Student and mentor gender are both associated with differences in rates of student’s continuation to positions that allow formal academic mentorship. Although trainees of women mentors are less likely to take on positions as academic mentors than trainees of men mentors, this effect is reduced substantially after controlling for several measurements of mentor status. Thus, the effect of mentor gender can be explained at least partially by gender disparities in social and financial resources available to mentors. Because trainees and mentors tend to be of the same gender, this association between mentor gender and academic continuation disproportionately impacts women trainees. On average, gender homophily in graduate training is unrelated to mentor status. A notable exception to this trend is the special case of scientists having been granted an outstanding distinction, evidenced by membership in the National Academy of Sciences, being a grantee of the Howard Hughes Medical Institute, or having been awarded the Nobel Prize. This group of mentors trains men graduate students at higher rates than their most successful colleagues. These results suggest that, in addition to other factors that limit career choices for women trainees, gender inequities in mentors’ access to resources and prestige contribute to women’s attrition from independent research positions.

An imbalance in gender representation has been identified across many fields of academic research. Men are disproportionately represented, especially at more senior career stages. Efforts to increase recruitment of women into training programs has increased their representation in many fields, but a "leaky pipeline" is still observed: a disproportionate number of women leave research before obtaining an independent position. To understand factors that lead to attrition of women from academic research, we analyzed a database of mentoring relationships covering the period 2000 to 2015 in several science and humanities fields. While representation of women has increased over time among both trainees and independent researchers, male graduate students and postdocs still continue to independent research at higher rates than their women colleagues. We analyzed aspects of the mentoring relationship that contribute to this continuing bias. One strong pattern that is observed is homophily, where trainees and mentors tend to be of the same gender. It has been hypothesized that institutional bias may limit the influence of female investigators, which could in turn impact the retention of their trainees in academia. Indeed, we observe trainees of female mentors are less likely to continue into independent research careers than trainees of male mentors. This disparity persists, but is reduced, after controlling for temporal trends in retention rates and under-representation of women mentors at the highest levels of funding and citation. Although the trainee pool for mentors that receive highly prestigious awards is skewed towards men, homophily is otherwise unrelated to mentor achievement. These results suggest that, in addition to other factors that limit career choices for female trainees, institutional bias that gives female mentors access to fewer resources contributes to the leaky pipeline for gender.