PaperlayerView exact revision on arXiv

Source-linked AI summary

Gender-Based Homophily in Research: A Large-Scale Study of Man-Woman Collaboration

Marek Kwiek, Wojciech Roszka

arXiv:2006.03935v4cs.DL

TL;DR

The paper examines gender patterns in research collaboration, an area rarely studied in Polish science. It analyzes Polish professors’ publication portfolios and finds that same-sex collaboration is substantially more characteristic of men than women, with patterns varying across age, discipline, and institutional research intensity.

  • Problem

    Gender collaboration patterns, including homophily, had rarely been studied in Polish science.

    Source: §2.4 · p. 4

  • Method

    The authors constructed individual publication portfolios for Polish professors and used descriptive analyses and fractional logit regression to examine same-sex collaboration ratios.

    Source: §3.2 · p. 8, §4.6 · p. 18, §5 · p. 21

  • Results

    Male scientists’ same-sex collaboration ratio is more than three times that of female scientists, and regression estimates it is 20 percentage points higher for men.

    Source: §5 · p. 21

  • Takeaways & Limitations

    Across the studied age groups, female scientists tend to collaborate with men, while male scientists also tend to collaborate with men.

    Source: §5 · p. 21

    Because 2017 characteristics are combined with decade-level publication averages, longitudinal and cohort analyses were not possible.

    Source: §3.1.1 · p. 7

Abstract

from arXiv · show

We examined the male-female collaboration practices of all internationally visible Polish university professors (N = 25,463) based on their Scopus-indexed publications from 2009-2018 (158,743 journal articles). We merged a national registry of 99,935 scientists (with full administrative and biographical data) with the Scopus publication database, using probabilistic and deterministic record linkage. Our unique biographical, administrative, publication, and citation database (The Polish Science Observatory) included all professors with at least a doctoral degree employed in 85 research-involved universities. We determined what we term an individual publication portfolio for every professor, and we examined the respective impacts of biological age, academic position, academic discipline, average journal prestige, and type of institution on the same-sex collaboration ratio. The gender homophily principle (publishing predominantly with scientists of the same sex) was found to apply to male scientists - but not to females. The majority of male scientists collaborate solely with males; most female scientists, in contrast, do not collaborate with females at all. Across all age groups studied, all-female collaboration is marginal, while all-male collaboration is pervasive. Gender homophily in research-intensive institutions proved stronger for males than for females. Finally, we used a multi-dimensional fractional logit regression model to estimate the impact of gender and other individual-level and institutional-level independent variables on gender homophily in research collaboration.

1. Introduction

The paper tests gender homophily in research collaboration: whether scientists predominantly publish with same-sex colleagues. It examines individual scientists’ publication portfolios across six research and institutional lenses in internationally visible Polish academia.

1. Introduction

  • The study compares male–male, female–female, and mixed-sex research collaboration, testing the prevailing view that scientists tend to collaborate with others of the same sex.The paper frames co-authorship patterns as a test of the homophily principle.Source: §1 · pp. 1–2
  • Homophily may simplify communication while also limiting people’s social worlds, and research collaboration offers a setting to examine this principle.The passage links the potentially growing influence of homophily to increasingly collaborative science.Source: §1 · pp. 1–2
  • The analysis covers 25,463 Polish professors and 158,743 Scopus-indexed publications from 2009–2018, examining patterns across six research and institutional lenses.The scientist, not the article, is the unit of analysis.Source: §1 · p. 2
  • An individual publication portfolio records each scientist’s distribution of same-sex and mixed-sex collaboration types, allowing the analysis to examine collaboration at the individual level.The study generalizes its individual-level results to Poland’s national higher education system.Source: §1 · p. 2

2. Literature review

The literature review situates gender collaboration within broader inequalities in science and competing explanations of researchers’ choices. It defines gender homophily, reviews its possible career implications, and identifies limited study of gender collaboration patterns in Polish science.

2.1. The gender context of science

  • Polish women form a substantial, productive, and internationalized academic workforce, yet prior research reports gender disparities in academic positions, collaboration, recognition, and career outcomes.The review notes that women made up 41.5% of the Polish university professors in the study sample.Source: §2.1 · pp. 2–3

2.2. Female scientists and competition

  • Competition-based accounts suggest gender differences in willingness to enter competitive settings may shape journal choices and team formation, while women in male-dominated disciplines may face added performance pressure.The review describes possible differences in confidence and preferences, and notes women may feel especially visible in male-dominated fields.Source: §2.2 · p. 3

2.3. Gender homophily in research collaboration defined

  • Gender homophily means disproportionate same-gender co-authorship; it may simplify collaboration but also exclude women from informal networks, and prior findings about which gender shows more homophily vary.The review notes that collaboration patterns differ by discipline and across countries.Source: §2.3 · p. 4

2.4. Hypotheses of this research

  • Gender collaboration patterns in Polish science had rarely been studied, motivating seven research questions and hypotheses about the topic.The review distinguishes this gap from prior work on Polish international collaboration and on productivity and impact.Source: §2.4 · p. 4

3. Data and methods

The study integrates Polish administrative and biographical records with Scopus publication data to construct a researcher-level dataset and examine collaboration patterns across the 2009–2018 period. Its sample and temporal comparisons are bounded by Scopus coverage, collaborator-gender estimation, and decade-aggregated measures.

3.1. Dataset

  • The integrated database selected 54,448 of 99,535 registered scientists at 85 universities, then retained 158,743 journal articles from 2009–2018 for analysis.Scopus metadata included 384,736 publications before journal-article selection.Source: §3.1 · pp. 5–6
  • Probabilistic name matching and clerical review produced 32,937 unique publication authors, including 25,463 journal-article authors.Name pairs were compared within universities, and posterior match probabilities above 0.85 were accepted; incorrectly matched records were deleted.Source: §3.1 · pp. 5–6

3.1.1. Limitations

  • The sample includes only internationally visible authors with Scopus-indexed publications, and other databases, periods, formats, or languages might yield different results.The authors identify this selection as a possible source of bias.Source: §3.1.1 · p. 6
  • Year-by-year and cohort analyses were not possible because age and position were measured in 2017 while publication portfolios and several other measures aggregate 2009–2018.The authors use a decade of output because Polish scientists publish too little annually for a single-year Scopus overview.Source: §3.1.1 · pp. 6–7

3.2. Methods

  • GenderizeR estimated gender for 7,640,123 OECD authors with probability at least 0.85, while gender was determined for 83,702 of 164,908 international collaborators.The reference database excluded non-research-intensive OECD and non-OECD universities, limiting collaborator coverage.Source: §3.2 · pp. 7–8
  • The same-sex collaboration ratio measures the same-sex share of a scientist’s collaborative articles, while dominant discipline is the most frequent ASJC category.Disciplines with fewer than 50 scientists were excluded; the ratio does not adjust for the gender composition of each discipline, and the authors note cross-disciplinary collaboration.Source: §3.2 · pp. 7–9

3.3. Sample

  • The sample comprised 25,463 scientists—14,886 men and 10,577 women—with women especially underrepresented in computer science, engineering, physics and astronomy, and mathematics.The respective female shares were 16.5%, 14.9%, 16.6%, and 25.2%.Source: §3.3 · p. 9

4. Results

The results examine same-sex collaboration among Polish scientists, including how patterns vary across age, academic position, discipline, institution type, and publication collaboration type. They also assess how individual and institutional characteristics relate to same-sex collaboration using a fractional logit model.

4.1. The same-sex collaboration ratio by gender

  • Males’ median same-sex collaboration ratio was 0.500, compared with 0.153 for females, so Hypothesis 1 was not confirmed.The national-sample median was 0.333, and the gender difference was significant at 0.05.Source: §4.1 · p. 9

4.2. The same-sex collaboration ratio by age and academic position

  • With age, males’ median same-sex ratio fell by 6 percentage points while females’ rose by 18, yet females’ ratio remained lower.By academic position, males’ ratio declined while females’ increased two and a half times, from 0.105 for assistant professors to 0.250 for full professors; age and position were interdependent across all 24 disciplines.Source: §4.2 · pp. 10–12

4.3. The same-sex collaboration ratio by academic discipline

  • The mean same-sex ratio rose weakly with the male share of a discipline (r = 0.228, R 2 = 0.052), while discipline-specific gender patterns varied widely.In COMP, ENG, and MATH, at least half of males collaborated only with males; in COMP, ENER, ENG, HEALTH, PHYS, and VET, at least half of females did not collaborate with females.Source: §4.3 · p. 12

4.4. The same-sex collaboration ratio by institutional type

  • In research-intensive IDUB institutions, male same-sex collaboration was higher than the system-wide level, while female same-sex collaboration was lower.Gender differences in median ratios were larger in IDUB institutions, where both male and female scientists were more likely to collaborate with men; the hypothesis was confirmed for males but not females.Source: §4.4 · pp. 13, 15, 16

4.5. The same-sex collaboration ratio by journal prestige

  • Mixed-sex articles appeared in more prestigious journals on average than same-sex articles, and solo articles had the lowest prestige.Median prestige for all-male and all-female publications was 59.17 and 58.00, respectively; prestige differences by gender were marginal for mixed-sex work but substantial for same-sex work, favoring male-only collaborations.Source: §4.5 · pp. 17–18

4.6. A modeling approach: A fractional logit regression model

  • The authors introduce fractional logit regression for the same-sex collaboration ratio, a zero-to-one outcome, with scientists modeled within institution and discipline contexts.The model includes individual and institutional factors, discipline characteristics, and mean individual publication prestige; the supplied passages provide no regression findings.Source: §4.6 · p. 18
  • Being male raises the same-sex collaboration ratio by 0.2, publishing in a male-dominated discipline by 0.167, and working in an IDUB institution by 0.048.A ten-year age increase raises the ratio by 0.01 on average; gender and male-dominated disciplines have the greatest influence.Source: §4.6 · p. 20
  • Being a full professor lowers the ratio by 5.6 percentage points, an associate professor by 1.4 points, and publishing mainly in STEM by 3.9 points.Journal prestige and institutional employee count are the weakest, though still significant, predictors.Source: §4.6 · p. 20
  • The model rejects Hypothesis 7: both individual- and institutional-level predictors affect gender homophily, with being male and working in a male-dominated discipline most influential.Being male is as influential as working in a male-dominated discipline and being employed in a research-intensive university.Source: §4.6 · p. 20
  • The model has R^2 = 0.159, suggesting other same-sex collaboration predictors are not captured by the available data.The residuals were non-normal, so robust standard errors were used; no influential observations were found because standardized residuals stayed within ±3 standard deviations.Source: §4.6 · pp. 20–21

5. Summary of findings, discussion, and conclusions

In Polish science, male scientists show substantially more same-sex collaboration than female scientists, and all-male collaboration is pervasive while all-female collaboration is marginal. The pattern varies with age, career stage, discipline, and institutional research intensity; the study combines broad national coverage with individual-level analysis.

Data and contribution

  • The study combines biographical and administrative data with bibliometric variables to analyze scientists’ publication portfolios across multiple dimensions.The sample includes 25,463 professors at 85 universities and 158,743 Scopus-indexed articles; it also uses biological age and academic stages rather than a first-publication proxy, alongside discipline, journal prestige, and institution type.Source: §5 · p. 21

Main findings

  • The same-sex collaboration ratio for males is more than three times that for females, with all-male collaboration pervasive and all-female collaboration marginal.The fractional logit analysis confirms a 20-percentage-point higher ratio for males.Source: §5 · p. 21

Journal prestige

  • Mixed-sex collaboration articles are, on average, published in more prestigious journals than same-sex collaboration articles.The comparison is based on average journal prestige across individual publication portfolios.Source: §5 · p. 21

Age and academic position

  • Across age groups and academic positions, males tend to collaborate with males, while females tend not to collaborate with other females.For young scientists, half of males have at least 54% of papers coauthored with males, compared with 6.3% of papers coauthored with females for half of young women; the median female ratio rises to 0.24 among older women from 0.06 among young women.Source: §5 · p. 22
  • The year-by-year analysis shows a downward same-sex collaboration ratio for males and an upward trend for females.In Poland, age and academic position are strongly correlated.Source: §5 · p. 22

Academic discipline

  • As the female share of a discipline rises, women increasingly coauthor with women; in computer science, engineering, and mathematics, at least half of men collaborate exclusively with men.In computer science, engineering, health professions, and physics and astronomy, at least half of women do not collaborate with women; in several female-dominated fields, female median same-sex collaboration exceeds the male median.Source: §5 · pp. 22–23

Institutional research intensity

  • Research-intensive institutions have higher all-male collaboration proportions and lower all-female collaboration proportions than other institutions.The comparison covers 10 research-intensive institutions and 75 other institutions.Source: §5 · p. 23

Fractional logistic regression

  • Being male and working in a male-dominated discipline are the strongest predictors of the same-sex collaboration ratio, followed by employment at a research-intensive university.The reported average marginal effects are 20 percentage points for male scientists, 16.7 percentage points for working in a male-dominated discipline, and 4.8 percentage points for working in a highly research-intensive institution; individual- and institutional-level factors both matter.Source: §5 · p. 23
Loading 2006.03935v4…