Gender bias in research teams and the underrepresentation of women in science

Why are females still underrepresented in science? The social factors that affect career choices and trajectories are thought to be important but are poorly understood. We analyzed author gender in a sample of >61,000 scientific articles in the biological sciences to evaluate the factors that shape the formation of research teams. We find that authorship teams are more gender-assorted than expected by chance, with excess homotypic assortment accounting for up to 7% of published articles. One possible mechanism that could explain gender assortment and broader patterns of female representation is that women may focus on different research topics than men (i.e., the \"topic preference\" hypothesis). An alternative hypothesis is that researchers may consciously or unconsciously prefer to work within same-gender teams (the \"gender homophily\" hypothesis). Using network analysis, we find no evidence to support the topic preference hypothesis, because the topics of female-authored articles are no more similar to each other than expected within the broader research landscape. Instead, consistent with a model of moderate gender homophily, we find that the prevalence of matched-gender teams increases as a discipline moves towards gender parity. This can occur because latent preferences are more easily fulfilled in a gender-diverse environment. Finally, we show that female authors pay a substantial citation cost to work in gender-matched teams. Notably, the prevalence of homotypic assortment is predicted to increase in the future if more disciplines shift towards gender parity. These data indicate that social preferences can have important downstream consequences for the retention of women and other underrepresented groups in science.