Gender-diverse research teams do better science – Study
The advantage gender-balanced teams have over all-men or all-women teams hold for both small and large teams. Mixed teams of six researchers are 7% more likely to produce novel papers and are 14.6% more likely to publish upper-tail (highly cited) or ‘home-run’ papers than are teams of the same size that are not gender diverse.
“Gender-diverse teams produce more novel and higher-impact scientific ideas,” says the paper authored by Professor Brian Uzzi, Richard L Thomas professor of leadership at the Kellogg School of Management at Northwestern University in the United States, and four other professors.
“There are numerous studies demonstrating the benefits of gender diversity in the workplace and in other business contexts,” says co-author Professor Yang Yang, who teaches at the Mendoza College of Business at the University of Notre Dame, Indiana.
“When we started looking at this in the realm of scientific research, we weren’t sure what we’d find. But our findings turned out to be very clear that gender diversity helps scientific novelty in medical research.”
The other co-authors of the study are Tanya Y Tian, New York University professor of management and organisations; Teresa K Woodruff, Michigan State University Foundation professor in the department of obstetrics, gynecology and reproductive biology (as well as being the university’s provost and executive vice-president for academic affairs) and Benjamin F Jones, Gordon and Llura Gund Family professor of entrepreneurship at the Kellogg School of Management, Northwestern University.
In an e-mail to University World News, Professor Donna Ginther, who is the Roy A Roberts and Regents distinguished professor of economics, a director of the Institute for Policy and Social Research at the University of Kansas, and a research associate at the National Bureau of Economic Research, called the study by Uzzi and his team “a tour de force of empirical work that demonstrates the benefit of mixed-gender teams in science”.
Under-representation of mixed-gender teams
Uzzi and his co-authors examined 6.6 million papers in medical science published since 2000. They found that the percentage of women scientists rose from 38% to 45%. Concomitantly, the percentage of papers by six or more authors rose from 30% to 50%.
As far back as 2000, the majority of groups in which there were four or more researchers were mixed-gender teams: 60% of the groups of four researchers were mixed gender in 2000, a percentage that rose to 70% in 2020. Over the same period, the percentage rise in mixed-gender groups of six or more researchers rose from 80% to 90%.
“Still,” writes Uzzi, “mixed-gender teams are significantly under-represented in medical science by up to 17% depending on team size when we measure team gender diversity.”
To determine whether a research paper is novel – in other words, does it combine knowledge in a new way relative to existing literature – Uzzi’s team employed a novelty measure developed in 2013.
This measure uses a paper’s listed references as an indicator of its mixture of knowledge. The measure considers papers with statistically atypical combinations of references to be novel because they create new combinations of knowledge that have not been, or rarely, joined in previous research.
“Papers whose bibliographies contain journal pairings that have frequently occurred together in the past (more than expected by chance) indicate the relatively conventional and familiar pairings,” writes Uzzi. By way of example, Yang told this writer to think of an article on economics that cites both the New York Times and the Wall Street Journal as being conventional and familiar.
By contrast, “papers whose bibliographies contain journal subject pairings that have been rarely co-cited represent novel combinations of knowledge”.
“However unusual one member of the combination may be,” says Uzzi, “it has credibility because it has passed peer review by other scientists who are familiar with the literature in the field.”
As part of what Yang called the “robustness test” of their data, the team considered a number of alternative explanations for why gender-diverse teams produced more novel research.
These factors include the team members’ network size, range, career age and international diversity as well as density of team members and expertise across team members.
(‘Density of team members’ refers to the proportion of member pairs who have collaborated before working on the paper in question while a team with members from computer science, economics and medicine would be considered as having higher expertise diversity than a team from, for example, economics alone.)
Uzzi and the other researchers found that when compared to same-gender teams, mixed-gender teams do have different expertise, network age diversity as well as different international diversity characteristics.
Mixed-gender teams are more likely to have higher topic-related expertise diversity, lower network density but larger network range and size. Additionally, mixed-gender teams are more likely to have a higher geographical diversity and internationalism and a higher degree of career age diversity.
After examining these differences through a regression model, Uzzi’s team found that there was statistically significant evidence to “indicate that gender diversity teams correlate with expertise diversity, network characteristics and demographic drivers of team success, which may inform the performance advantages seen among mixed-gender teams”.
Yet, these figures were not enough to explain the performance advantage of mixed-gender teams.
“What the data shows,” explains Uzzi, “is that if you have a same-gender team and a mixed-gender team with the same level of expertise, diversity and the same network characteristics, the mixed-gender team would still outperform the same-gender team. This indicates that the mixture of genders does something on top of these other factors to produce high-performance science.”
Women are not ‘the weak link’
The study also considered the impact of the little-discussed research phenomenon known as ‘gender homophily’, the citing by male researchers of papers by men and vice versa by women. Far from finding evidence of homophily, Uzzi’s team found, “mixed-gender teams receive more citations than same-gender teams regardless of the citing team”.
In other words, all-women, all-men and mixed-gender teams cited research generated by mixed-gender teams more often than they did research generated by single-gender teams.
Further, Uzzi says, the performance of mixed-gender teams increases the closer the teams are to gender parity.
“This is important because there has been talk in the literature and media that when you put men and women together, maybe the team does better but women are still ‘the weak link’. What the greater balance, the greater outcome, shows is that women are not ‘the weak link’ because if they were, then you would expect the team that was majority men to outperform. And that is not what the data shows.”
What the study did not set out to determine is what mechanisms cause gender-diverse teams to outperform same-gender teams and what policies could be put in place to foster the creation of these teams. With caveats in place, Uzzi proposed a number of questions that indicate the direction of his thinking about what processes within gender-diverse teams give them that “extra lift”.
These include: How do people exchange information? How is respect shown towards each other on a team? How much ‘turn taking’ do they do when exchanging ideas? What information do they actually exchange?
Uzzi’s questions point to the type of inquiry, Ginther told University World News, that should come next. “Subsequent research should use author contribution statements to unpack the mechanisms that give rise to the increased novelty and impact of gender-diverse teams.”
A similar pattern across all disciplines
The story Uzzi told this writer about the history of their study is a textbook case of how science works. One of the reviewers to whom they sent a draft of the paper asked if the findings were particular to medical science.
The reason the reviewer asked is because a lot of medical science is gendered. You’ve got products for women or men and treatments for women or men because of biological differences. Perhaps having an extended team allows you to make products that are more comprehensive across the genders, says Uzzi.
To test their findings in medical science, Uzzi’s team expanded their database from 6.6 million to 26 million articles in a number of fields.
“We went to all the other disciplines in science and did the same analysis. We looked at physics, chemistry, biology, sociology, economics and other fields. And we found the same pattern.
“And that’s what really makes the paper so massive. Once we do this in a number of disciplines, it [the finding that gender-diverse teams produce more novel and higher-impact science] begins to feel very much like a fact.”