Physics research papers authored by women receive fewer citations than expected given the number of women who work in the field and the number of papers they publish, according to a new analysis.
The study, posted in December on arXiv, explores citations in more than a million papers published in 35 high-impact physics journals. It found that the under-citation of papers coauthored by women relative to those by men is a consistent trend across eight subfields of physics.
Gender disparities in citations have already been documented in many other fields, including astronomy, neuroscience, economics, and medicine. “We were curious whether the same thing would hold in physics,” says Dani Bassett, a physicist and systems neuroscientist at the University of Pennsylvania in Philadelphia, who coauthored the new analysis.
Bassett and her colleagues downloaded all the papers published in the 35 journals from 1995 to 2020 that were available in the Web of Science database. They studied citation behavior by matching references’ digital object identifiers (unique IDs of scholarly papers) to those of other papers in the sample. Then the researchers used a model to predict papers’ expected citation rates based on characteristics other than author gender. Finally, they determined the number of citations for papers with a woman as first author, last author, or both and compared those numbers with the model’s gender-independent predictions.
The number of papers authored by women in the eight physics subfields examined in the study almost doubled between 1995 and 2020, from around 17% to roughly 33%, as shown in the graph above. But those manuscripts attracted about 3% fewer citations than expected, whereas those whose first and last authors were men were cited roughly 1% more.
What’s more, the gender gap was largest in papers authored by men. According to the study, manuscripts with male authors cited recent male-authored papers about 2% more than expected and cited recent papers authored by women 6% less. Studies with a female author over-cited recent female-authored papers by 3% and under-cited recent papers by men by 1%.
Journals grouped into the general physics category—which includes many of the highest-impact physics titles—had among the smallest percentage of manuscripts with women as first authors, last authors, or both throughout the 26-year period. Astronomy and astrophysics, on the other hand, contained the highest proportion of papers authored by women. The citation gap roughly tracked the overall representation of women: As shown in the graph to the right, general category titles had the widest gap (gray), whereas astronomy and astrophysics had the narrowest (pink).
The gender citation gap persisted even though Bassett and her colleagues’ model accounted for the seniority of the authors, as senior faculty tend to attract more citations, and for the age of the papers, as older papers have had more time to accumulate citations than newer ones. The researchers also controlled for the type of study, because review articles often garner more citations than empirical studies, and for the number of coauthors, because studies with more authors have been shown to attract more citations.
The most likely reason for the citation gap, Bassett says, is that physicists “implicitly associate a woman’s name with less meritorious scientific work.” When scanning articles and deciding what to read or cite, readers may also unconsciously give more weight to authors from certain institutions, locations, and races, she notes.
Rachel Werner, a health economist at the University of Pennsylvania who has studied gender biases in medicine, notes that the new study also found that disparities are largest when authors are citing papers in fields outside their own. “That supports the idea that this is an implicit bias,” she says. “Often when people are faced with uncertainty, they fall back on heuristics.”
Another issue, Bassett notes, is that scientists often gain visibility from publishing in well-known journals, whose author demographics may not match those of the research community. Physicists may also be more likely to cite people they know or those who give talks and present their work at conferences, she adds. “The majority of keynote talks are being given by a single race and ethnicity: a white man.”
The new study’s methodology is solid, says Laura McCullough, a physicist at the University of Wisconsin–Stout, who has written about gender biases in physics. Neven Caplar, a scientific software developer at Princeton University in New Jersey who coauthored the study about gender disparities in citations in astronomy, agrees.
One limitation of the study is that it couldn’t decipher the gender of about one-fifth of the authors, those who list only their initials instead of their first names, Bassett notes. Although Bassett says she and her team excluded those authors from their sample, McCullough thinks a significant number of them could be women. She says women in science often hide their first names to avoid discrimination.
Another problem, Bassett says, is that the software determines the chance of an author being a certain gender on the basis of his or her name, but it will be wrong at least some of the time, especially for gender-neutral names. It also cannot identify nonbinary individuals.
The study suggests potential solutions for citation disparities. For instance, Bassett evaluates her own papers on how diverse the references are in terms of gender, race, and ethnicity. “I look at them, and I statistically assess” whether the proportions are consistent with the field that the paper is from, she says. “If they are not consistent, I go back to the literature, and I use it as an opportunity for self-education.”
Bassett also includes a citation diversity statement in her papers. The short paragraph’s purpose is to explain any potential biases in the citations, to break down the diversity of the citations, and to raise awareness about citation biases. One study has found that papers with such diversity statements are more likely to have citation practices closer to expected rates. More journals could encourage authors to include such statements in submission forms or on journal pages, Bassett says. They could also actively try to publish more papers by women, invite more women to be peer reviewers, or ensure gender equality in special issues, she adds.
Journals could also remove limitations on the length of reference lists, Bassett notes. “What we see in our data is that the longer the reference list, the more likely that list is to be equitable across genders.”
