Citation inequity and gendered citation practices in contemporary physics

I. INTRODUCTION

• Under-attribution of women’s scholarship has a long history and is associated with devaluation, fewer awards, longer review processes, and fewer invited commissions.The authors frame this as a consequentialist study of the effects of current practices rather than assuming individual intent.Source: §I · pp. 1–2
• The study quantifies name-based under-citation in physics and finds that citation imbalance varies by citing author, venue, and citation form.The authors use a proxy based on statistical correlations between publishing forenames and gender identity.Source: §I · p. 2

II. METHODS

• The analysis covers over one million papers from 35 physics journals and examines citation practices by author category, subfield, and publication date.The introduction describes the broader methods as building on prior research on gendered citation bias.Source: §I · p. 2, §II · p. 2

I. INTRODUCTION

• The authors recommend thoughtful attention to reference-list gender composition and citation diversity statements, especially when citing work outside one’s immediate knowledge or network.They also report that longer reference lists tend to show less over-citation of man-authored papers, and suggest journals reconsider reference-list limits.Source: §I · p. 2

1. Journal selection

• The journal sample comprised 35 journals in seven non-overlapping subfield lists, chosen to cover major physics areas and represent each subfield equally by journal count.Within each subfield, central journals were selected using 2018 Eigenfactor scores.Source: §1 · p. 2

2. Data collection

• The authors downloaded papers published in the selected journals from 1995 to 2020 and retained original research and review articles.They extracted author names, references, publication dates, and DOIs.Source: §2 · p. 2
• Citation links were identified by matching DOIs in reference lists to DOIs of papers in the dataset; missing first names were sought through Crossref and name disambiguation.First names were needed for the study’s name-based author gender categorization.Source: §2 · p. 3

1. Author name disambiguation

• Name disambiguation assigned full first names to 88,780 of 463,538 initials-only entries, a success rate of approximately 19.15%.Ambiguous initials were left unchanged; name variants were standardized when matching names had no conflicting initials.Source: §1 · p. 3

2. Estimation of publication month

• For papers lacking publication month, the midpoint between estimated bounds had an average absolute error of approximately 2.27 months, versus approximately 2.98 months for guessing June or July.The test used the 1,028,853 papers with known publication months.Source: §2 · p. 3

3. Name-based assignment of author gender categories

• Author categories were assigned from first and last names using a probability threshold of at least 0.7, and papers with at least one woman-labeled endpoint author were grouped as W||W.Both first and last authors received labels in 60% of papers; the authors stress that these categories proxy perceived gender associations in names, not necessarily true gender identities.Source: §3 · pp. 3–4

4. Reference list cleaning procedure

• The cleaned reference lists excluded self-citations, citations outside the dataset, and citations to papers whose authors could not be gender-categorized.Self-citations were defined conservatively using first- or last-author overlap between citing and cited papers.Source: §4 · p. 4

1. Probability estimation of author gender category according to paper characteristics

• A generalized additive model predicted MM or W||W authorship from publication timing, author publication counts, author count, journal, and article type.The model used no interactions between these predictors.Source: §1 · p. 4
• The model’s predicted probabilities supplied expected MM and W||W citation counts under gender-agnostic selection among papers with comparable characteristics.Observed citations were compared with these expectations to quantify over- or under-citation.Source: §1 · pp. 4–5

2. Calculation of over-/under-citation

• Over-/under-citation is the percent difference between observed citations and gender-blind expected citations.Expected MM citations are summed from model-estimated probabilities for each citation.Source: §2 · p. 5

3. Subfield delineation

• Subfield boundaries combine predefined journal categories with post-hoc citation-network clustering.The subfields and their constituent journals are shown in Fig. 1b.Source: §3 · p. 5

A. Time-varying demographics of published papers

• Woman-authored papers rose from 17% of MM and W||W papers in 1995 to 33% in 2020, but representation remained uneven across subfields and journals.General and high-energy physics had the lowest W||W proportions, while astronomy and astrophysics generally had the highest.Source: A. Time-varying demographics of… · p. 5

B. Citation imbalance exists & varies by citing venue

• Most journals fall on the MM-overcitation side of the comparison, while only three significantly prefer citing W||W papers.The three are Journal of Nuclear Materials, Astronomy & Astrophysics, and Soft Matter.Source: B. Citation imbalance exists & varies… · p. 6

C. Citation imbalance varies by citing actor

• MM citing teams over-cited MM papers by 2.05% and under-cited W||W papers by 6.53%, while W||W teams favored W||W papers by 4.94%.The corresponding W||W-team pattern was 3.56% over-citation of W||W papers and 1.38% under-citation of MM papers.Source: C. Citation imbalance varies by citing… · p. 6

B. Citation imbalance exists & varies by citing venue

• MM papers were cited about 1.06% more than expected and W||W papers about 3.17% less, for an overall gap of roughly 4.23%.The gap varied by subfield and journal; general physics had the widest collective gap, while astronomy and astrophysics had the narrowest.Source: B. Citation imbalance exists & varies… · p. 5, Figure 2 · p. 7

C. Citation imbalance varies by citing actor

• The largest citing-team gaps occurred in general physics: 16.47% in favor of MM papers for MM teams and 8.8% in favor of MM papers for W||W teams.W||W teams in astronomy/astrophysics and soft matter/biophysics instead showed preferences for W||W papers, with gaps of 8.42% and 6.53%.Source: C. Citation imbalance varies by citing… · p. 7

D. Stable and growing trends in citation imbalance over time

• The gap between MM over-citation and W||W under-citation remained relatively stable or slightly grew from 2009 to 2020.MM citing teams' gap widened, while W||W citing teams moved toward citation parity.Source: D. Stable and growing trends in… · pp. 7, 9, Figure 3 · p. 8
• MM citers' gap increased over time, whereas W||W citers' preference for W||W papers decreased toward parity.The trends contrast despite the growing fraction of published W||W papers over the same period.Source: D. Stable and growing trends in… · p. 9
• In 2020, the citation gap reached 16.67% in general physics and 10.7% in condensed matter.These were the two subfields whose gaps increased between 2009 and 2020.Source: D. Stable and growing trends in… · p. 9

E. Citation imbalance varies according to its form

• Unfamiliar references show larger citation gaps: out-of-subfield MM papers are over-cited 1.90% and W||W papers under-cited -7.36%, versus 0.66% and -1.80% within-subfield.For unshared-author papers, MM over-citation is 1.29% and W||W under-citation is -4.34%, versus 0.57% and -1.39% for shared-author papers.Source: E. Citation imbalance varies according… · p. 9
• For familiar citations, MM teams preferentially cite MM papers and W||W teams preferentially cite W||W papers, producing competing effects on the overall citation gap.These tendencies may be partly explained by differences in the local co-authorship networks around MM and W||W papers.Source: E. Citation imbalance varies according… · p. 9
• For unfamiliar citations, W||W teams cite MM and W||W papers approximately equitably, while MM teams retain a preference for MM papers.MM teams’ preference is enhanced for out-of-subfield citations and remains significant, though slightly reduced, for unshared-author citations.Source: E. Citation imbalance varies according… · p. 9
• Both familiarity proxies compare whether citing and cited papers share a subfield or an author; the overall gap is more marked for unfamiliar citations because citing-team behaviors differ.Figure 4 shows these comparisons, with 95% confidence intervals estimated from 500 bootstrap resampling iterations.Source: E. Citation imbalance varies according… · p. 9, Figure 4 · p. 10

F. Additional correlates of citation imbalance

• Across 35 journals, a higher fraction of W||W papers generally correlates with greater citation preference for W||W papers.The authors characterize representation within journal authorship pools as a meaningful correlate, not a demonstrated cause, of citation imbalance.Source: F. Additional correlates of citation… · p. 10, Figure 5 · p. 11
• Longer reference lists tend to show greater citation preference for W||W papers across both MM and W||W citing teams, a weak but statistically significant collective trend.The relationship is stable over time; averaged slopes suggest about a 1.5% increase in W||W citation preference for every 10 additional citations.Source: F. Additional correlates of citation… · p. 10, Figure 5 · p. 11

IV. DISCUSSION

• The study finds global over-citation of papers in the man-author category and under-citation of papers in the woman-author category, with imbalance varying by citing author, journal, familiarity, authorship representation, and reference-list length.The analysis spans 35 physics journals over 25 years and estimates citation imbalance while controlling for paper characteristics.Source: §IV · p. 11

A. Homophilic behavior and its drivers

• MM citing teams preferentially cite MM papers, while W||W citing teams preferentially cite W||W papers, relative to the model’s predictions.MM homophily persists over time, across subfields, and for familiar and less familiar citations; W||W homophily weakens over time, varies by subfield, and is significant only for familiar citations.Source: A. Homophilic behavior and its drivers · p. 11
• The authors urge all physics citers to critically examine reference lists in light of historical and contemporary imbalances; women’s citation of women may reflect activist choice homophily and epistemic resistance.The paper presents activist choice as a possible driver, not an established explanation.Source: A. Homophilic behavior and its drivers · p. 12

B. Toward citation equity

• The authors highlight self-education about work in general physics and less familiar fields, and note citation diversity statements as an individual mitigation approach.They also describe journal responses including policies and tools; a retrospective analysis reported no statistically detectable gender disparities in papers with citation diversity statements.Source: B. Toward citation equity · p. 12
• The authors propose testing whether increasing women’s authorship in journals reduces citation imbalance, based on a noncausal correlation between women’s representation and citation patterns.They also suggest removing reference-list length limits may benefit gender equality, but explicitly reject implying that man-authored papers are more deserving when references are limited.Source: B. Toward citation equity · pp. 12–13

• Citations can be formal references or informal spoken and written acts that call forward particular scholars and scholarship.Source: C. Expanding the scope of equitable… · p. 13
• In writing, researchers can examine whose work shapes ideas, methods, and interpretation, and whether women’s citations are central or merely supplementary.Source: C. Expanding the scope of equitable… · p. 13
• In speaking and professional life, equitable representation can be considered in teaching, talks, recommendations, invitations, and collaboration.Source: C. Expanding the scope of equitable… · p. 13
• The authors invite further research and reflection on how citation imbalances and related dynamics appear beyond reference lists and how they might be addressed.Source: C. Expanding the scope of equitable… · p. 13

D. Methodological considerations

• Name-based assessments cannot reveal discrimination specific to trans and nonbinary people, and future work should examine citation practices across intersecting dimensions of identity.Source: D. Methodological considerations · p. 13

V. CONCLUSION

• Across 35 physics journals, citation practices skew toward over-citation of man-authored papers and under-citation of woman-authored papers, with behavior varying by citing team, venue, and familiarity.Source: §V · pp. 13–14

VI. CITATION DIVERSITY STATEMENT

• The citation diversity statement reports four name-based first/last-author gender categories and cautions that names, pronouns, and social-media profiles may not indicate gender identity.It also notes that probabilistic name studies cannot detect citation costs specific to intersex, nonbinary, or transgender people who are out to many colleagues.Source: §VI · p. 14