Women are credited less in science than men

A list of authors records who received credit, not everyone who helped produce knowledge. Matthew B. Ross and five co-authors make that distinction measurable in this 2022 study. Rather than starting only with published names, they identify people paid to work in research teams and ask who appears on those teams’ articles and patents. The resulting gender gap challenges the habit of treating publication counts as a direct window onto productivity. Some apparent differences in output may arise in the allocation of recognition itself.

Build a denominator beyond the byline

The administrative analysis links 128,859 people in 9,778 teams at 20 US universities, using employment records from 2013–2016 and scientific outputs from 2014–2016. It connects teams to 39,426 articles and 7,675 patents. These are the analytical sample, not every institution in the larger data infrastructure or a representative census of all research. Payroll records make workers visible who would disappear if an inquiry included only named authors.

Team membership nevertheless does not prove that each member contributed enough to deserve every byline. The authors construct potential authorship observations using roles, employment timing and weights for people associated with multiple teams and documents. Actual credit is compared with that constructed opportunity. This is a method for investigating patterns in attribution, not a tribunal deciding each individual dispute.

The fully adjusted model reports women as 13.24% less likely to be named on articles, a difference of about 0.421 percentage points. The two numbers describe the same estimated gap on different scales; 13.24 percentage points would be a different claim. Controls for roles, time and teams reduce the disparity without removing it. Because gender was not randomized, the regressions do not by themselves establish one exclusive causal mechanism or an equal contribution on every project.

The 2023 author correction matters to interpretation. It replaces claims that gaps are “significant” across specified groups with the more limited wording that gaps remain, corrects the count of patent fields, and clarifies a weighted attribution rate. A broad pattern should not be inflated into a claim that every subgroup comparison independently meets the same statistical standard.

Recognition is organized through authority

A separate survey received 2,660 responses from researchers already connected to published work. Questions about exclusion, reasons for exclusion and contributions to recent articles have different responding subsets. Their results cannot be merged into one denominator. The survey offers another view of recognition, but willingness to respond and access to an existing publication record shape whose experiences it captures.

Open-ended responses and six follow-up interviews describe unclear rules, senior investigators’ discretion and the risks of asking for credit. Speaking up can be necessary and can also invite humiliation. This complicates advice that women should simply advocate more strongly for themselves. When the authority deciding authorship also influences employment and future opportunities, negotiation is not conducted between equals. These accounts illuminate possible processes; they do not establish how frequently each process occurs throughout science.

The administrative evidence also has boundaries. Unpaid work is absent when no time is charged to a grant, research-intensive universities shape the setting, and gender is inferred through name-based algorithms that do not represent nonbinary or fluid identities. The survey excludes people never credited on a publication. The very strategy that brings omitted workers into view therefore still has edges. Better measurement requires asking which omissions a new dataset can address and which it reproduces.

The authors suggest that lost credit can discourage careers, while treating fuller longitudinal investigation as a further task. That distinction should remain: an estimated attribution gap is not a measured lifetime career penalty. A practical reading asks who sets authorship criteria, when they become known, and whether contributors have a credible route to challenge their application. It does not turn a statistical pattern into a verdict on any unnamed laboratory.

Data Feminism provides a complementary framework through its attention to invisible labor and the authority to define categories. Here that question reaches the evidence used to evaluate scientists themselves. If recognition determines access to future resources, measuring only those already recognized can make an unequal process look like a neutral record of merit.