Stack of Nature journals

At Nature, we need to work harder to eliminate biases in our journalism.Credit: Nature

A consistent finding of researchers studying the news media is that women are quoted much less often than men. The Gender Gap Tracker (GGT), an automated system created by a team at Simon Fraser University in Burnaby, Canada, has tracked seven Canadian news sites since October 2018 and found that 71% of interviewees quoted in articles were men (F. T. Asr et al. PLoS ONE 16, e0245533; 2021).

Four other media agencies around the world worked with the GGT to follow 5 days of their own news coverage last November, and found that 73% of quotes were from men. And the Global Media Monitoring Project reported in March that in 2020, just 25% of news sources and subjects were women — although this was an increase from 17% in 1995, when the project began its work.

Two researchers in the United States have now analysed Nature’s journalism — and found similarly sobering results.

The researchers estimate that, in 2020, some 69% of the direct quotes (not including paraphrased comments) in Nature’s journalistic articles were from men. This is according to a software analysis of the gender of people quoted in more than 16,000 Nature News, Features and Careers articles between 2005 and 2020. Overall, the proportion of men being quoted in Nature’s journalism has been falling. It was around 80% before 2017, and 87% in 2005.

These findings are an important and welcome reminder of gender bias in journalism — a problem that Nature’s editors are striving to address. The numbers also show how software can be used by writers and editors to recognize biases, and that at Nature we need to work harder to eliminate them.

The analysis, posted before peer review (N. R. Davidson and C. S. Greene Preprint at bioRxiv https://doi.org/10.1101/2021.06.21.449261; 2021), was automated with software. The researchers first scraped articles written by journalists and published on nature.com. They then wrote code to pick out the names of people quoted by reporters, counting those whose quotes are enclosed in speech marks. Another algorithm called genderize.io was used to assign gender, a standard approach for large bibliometric studies.

The idea for the study was developed in consultation with Nature, but the authors, Natalie Davidson and Casey Greene, both computational biologists at the University of Colorado School of Medicine in Aurora, worked independently to design and conduct the study.

The duo found one exception to the main finding on gender bias — in Nature’s Careers features articles. Quotes from men and women appear in equal numbers in this section, which comprises reporting from journalists on different aspects of research careers.

The study does not assess all of Nature’s non-primary-research output; for example, content written by invited expert authors is not included. This category of content tends not to directly quote other individuals.

But over the past five years, Nature has started collecting data on gender diversity among the authors of such commissioned content. For example, last year, women comprised 58% of authors in Nature’s World View column — up from 35% in 2017, and from 18% in 2016. And, in 2020, women accounted for 34% of authors of News and Views articles — which explain and analyse new research — compared with 26% in 2017 and 12% in 2012. Another example of journalism is our photo-essay section, Where I Work, which profiles researchers in places where they study. This has featured 56% female scientists since its introduction in 2019.

Caveats and limitations

As with all studies, there are some caveats. Not all names could be analysed, and Davidson and Greene note that their software has a slight male bias when it comes to assigning gender to names. For instance, in a sample of articles from 2005 to 2015, it assigned 78% of quoted speakers as male, but the true number, when the authors checked, was 75%. It also cannot estimate non-binary gender.

To help contextualize their findings, the researchers consider various ways of measuring the overall proportion of women in academic research. The global science report by UNESCO, the United Nations science and education organization, was published earlier this month and puts this at 33% in 2018. By comparison, Davidson and Greene found that women made up around 20% of last-author and 25% of first-author positions on Nature papers; the ratios are about 25% and 37% in a wider selection of papers in Springer Nature journals.

Davidson and Greene also analysed what they call the ‘name origins’ of quoted interviewees in Nature’s journalism — a linguistic analysis that assigns names to broad regions of the world where a particular name is over-represented. The authors use an algorithm called NamePrism that excludes the United States, Canada and Australia because of the diversity of names in these countries.

This analysis suggests that Nature’s journalists tend to quote more researchers with names commonly used in English-speaking cultures, and fewer with names that the algorithm classifies as being of East Asian origin (including China, Singapore, Vietnam and other southeast Asian countries). It also indicates that this imbalance in name origins is greater than that seen in the names of last authors in Springer Nature research papers.

Nature’s journalism team has been making efforts to track and improve its representation of all under-represented groups, but previously this has not been a centralized effort. In the past year, however, the team has been developing and trialling a prototype system, with the goal — while abiding by data-privacy regulations — of collecting information on the gender, career stage and location of journalistic sources, expert authors and other contributors. We hope to be able to use this to establish and report a set of baseline figures, and then to improve on them. We are also working hard to include more voices from all groups that are under-represented in research.

More than half of Nature’s journalism team is female, but the overwhelming majority of its members are in Europe, the United States and Australia. We recognize that we need to strive harder to find diverse sources across the world.

Journalists, non-profit organizations and scientists have written excellent guides to diversifying sources, such as at the Open Notebook. They have also produced databases of diverse experts in many scientific fields. And, as personal testimonies from science reporters and other journalists attest, keeping track of the numbers, as Davidson and Greene are doing, is an essential part of this process, so we can be reminded of just how much more we need to do.