Researchers collaborate with same-gendered colleagues more often than expected across the life sciences

Evidence suggests that women in academia are hindered by conscious and unconscious biases, and often feel excluded from formal and informal opportunities for research collaboration. In addition to ensuring fairness and helping to redress gender imbalance in the academic workforce, increasing women’s access to collaboration could help scientific progress by drawing on more of the available human capital. Here, we test whether researchers tend to collaborate with same-gendered colleagues, using more stringent methods and a larger dataset than in past work. Our results reaffirm that researchers co-publish with colleagues of the same gender more often than expected by chance, and show that this ‘gender homophily’ is slightly stronger today than it was 10 years ago. Contrary to our expectations, we found no evidence that homophily is driven mostly by senior academics, and no evidence that homophily is stronger in fields where women are in the minority. Interestingly, journals with a high impact factor for their discipline tended to have comparatively low homophily, as predicted if mixed-gender teams produce better research. We discuss some potential causes of gender homophily in science.


Gender and coauthorship
Overall set of papers Male-biased subset, e.g.
-Papers on Surgery -Older papers -Papers from Japan -Papers on Nursing -Newer papers -Papers from Serbia The Wahlund effect Illusory preferences for same-gendered collaborators One paper with two women authors Figure 1: The Wahlund effect can make it appear as if authors publish with same-gendered colleagues disproportionately often, even if collaboration is completely random with respect to gender. Here, coloured circles represent male and female authors, and coauthors are linked with lines. Across the whole set of ten papers, there is an apparent excess of same-gender collaborations: there are six same-gender papers and only four mixed-gender papers, which is fewer than the 10 × 2 × 0.5 × 0.5 = 5 mixed-gender papers expected under the null hypothesis that authors assort randomly. However, within each subset, there is no evidence that authors prefer to publish with same-gendered individuals (if anything, this small dataset suggests gender heterophily). The Wahlund effect will tend to inflate the frequency of same-gender coauthorships whenever the data is composed of two or more disconnected subsets of literature with different author gender ratios; these subsets could be research disciplines, older versus newer papers, or papers from authors in different countries. The example countries and disciplines were selected based on data in [5]. showed statistically significant evidence of gender homophily (denoted by α > 0), and 1 showed statistically significant evidence of heterophily (α < 0), after false discovery rate correction.
In the stacked density plot, the white area shows the number of journals for which homophily was significantly stronger than expected under the null hypothesis (corrected p < 0.05), while the blue area shows all the remainder. Patterns were similar whether α was calculated for all authors, for first authors only, or for last authors only. Points in the right panel show α for individual journals. The coefficient of homophily (α ) was slightly less positive when calculated for two-author papers only, relative to papers with longer author lists. The individual points, whose distribution is summarised by the violin plots, correspond to individual journals. The larger white points show the mean for each group (and its 95% CIs), as calculated by a Bayesian meta-regression model accounting for repeated measures of α within journals, as well as the precision with which α was estimated.
Relationship between gender homophily and gender ratio 132 We next tested whether researchers are more or less likely to publish with same-gendered 133 colleagues in strongly gender-biased disciplines (e.g. Surgery or Nursing), relative to disciplines 134 with a comparatively gender-balanced workforce (e.g. Psychiatry). We found a positive, non-135 linear relationship between the overall gender ratio of all authors publishing in a particular 136 journal [5], and the estimated value of α for all authors and for first authors (Figure 4).

137
Journals with a balanced or female-biased author gender ratio tended to have higher α 138 (i.e. stronger homophily) than journals with a male-biased author gender ratio (GAM smooth 139 terms p < 0.001; Online Supplementary Material). The relationship was not statistically 140 significant when α was calculated for last authors (GAM, p = 0.142), though the trend 141 appeared similar (Figure 4).

142
Relationship between journal impact factor and gender homophily 143 We observed a noisy but statistically significant linear relationship between standardised 144 journal impact factor and α , such that journals with a high impact factor for their discipline 145 had weaker gender homophily than did journals with a low impact factor for their discipline Given that we cannot identify individual researchers or their career stages, we used a simple 162 model to derive the theoretical expectations for α when the gender ratio differs between career 163 stages (see Methods). As shown in Figure 6, we predict that α is expected to be non-zero, 164 even if collaborators are randomly selected with respect to gender, provided that there is There is a weakly positive, non-linear relationship between the gender ratio of authors publishing in a journal, and the coefficient of homophily (α ). Specifically, journals with 50% women authors or higher tended to have more same-sex coauthorships than did journals with predominantly men authors. This relationship held whether α was calculated for all authors, first authors only, or last authors only. A negative value on the x-axis denotes an excess of men authors, a positive value denotes an excess of women authors, and zero denotes gender parity (i.e. equal numbers of men and women). The lines were fitted using generalised additive models with the smoothing parameter k set to 3. : Journal impact factor (expressed relative to the average for the discipline) is negatively correlated with α . The relationship is noisy (R 2 = 0.043), but the results suggest that journals with strong homophily tend to have lower impact factors than journals with weak homophily in the same discipline.
a gender gap between career stages. The extent to which α deviates from zero depends on 166 the relative frequencies of collaboration within and between career stages (rows and columns 167 in Figure 6), and the size of the gender gap between stages (x-and y-axes in Figure 6).

168
When >50% of coauthor pairs comprise one early-career and one established researcher, we 169 expect gender heterophily (α < 0) whenever the gender ratio differs between career stages.

170
Conversely, when >50% of collaborations are between people at the same career stage, we 171 expect gender homophily (α > 0). In a few parameter spaces (shown in red; Figure 6), α was 172 quite high, and overlapped with the values that we estimated ( Figure 2).

173
Despite this overlap, Figure 6 suggests that our main conclusions (and those of other studies  Lastly, we note that if there is a gender gap between career stages and coauthorships between 184 early-career and established researchers comprise >50% of the total, then the baseline 185 expectation for α is actually less than zero (blue areas in Figure 6). Therefore, it is possible 186 that researchers preferentially assort with same-gendered collaborators even more strongly 187 than implied by our results, at least for certain journals or research disciplines.

189
We found evidence that researchers work with same-gendered coauthors more often than suspect the contribution of these uncontrolled artefacts to be minor, for four reasons: we 205 Figure 6: When the gender ratio of early-career researchers is not equal to the gender ratio among established researchers, the null expectation for α is not necessarily zero. Specifically, if most collaborations occur between career stages, there will be an excess of mixed-gender collaborations (α < 0, blue areas), while if most collaborator pairs comprise two people at the same career stage, there will be an excess of same-gender collaborations (α > 0, red areas). However, the conditions required for strong gender homophily (i.e. the red areas) are quite restrictive, making it unlikely that this issue can fully explain the homophily observed in our study. Additionally, in research disciplines where between-career stage collaboration is common and there is a shortage of women among established researchers (i.e. the blue areas), our study will underestimate the strength of gender homophily. Contour lines denote increments of 0.1.
inflation of α in highly multidisciplinary journals relative to specialised journals; restricting 207 the data by country yielded similar estimates of α ; and we used a simple model to show that 208 differences in gender ratio between career stages are unlikely to fully explain our results. On 209 balance, we believe the data suggest that it is likely that some researchers preferentially select show such a preference, or how much the strength of the preference varies between individuals. 212 We also note that even in a world in which collaboration was completely random with respect with gender-balanced co-authors. 217 We hypothesised that disciplines with a strongly skewed gender ratio might show the strongest 218 gender homophily, e.g. because being in the minority might increase people's motivation to 219 seek out same-gendered colleagues. Contrary to this hypothesis, we found no evidence that 220 gender homophily is restricted to particular disciplines: α was similarly high across the board 221 ( Figure 2). Interestingly, gender homophily was weakest for journals with a male-biased 222 author gender ratio, and strongest in journals with a female-biased author gender ratio. One 223 possible reason is that men are more likely to preferentially seek out male collaborators in 224 fields where men are a minority, relative to the homophily displayed by women in fields where 225 women are a minority. However, this latter result only has tentative statistical support since 226 our sample contains few journals in which most authors are women (Figure 4). 227 We also found that gender homophily was marginally stronger in 2015-2016 relative to  academics. This finding could contribute to the homophily we observed, and is cause for 266 concern since the results might reflect discrimination against women during hiring [20], or 267 avoidance by women of elite research groups (e.g. due to gender differences in confidence, 268 or a perception that some groups are sexist). We also found a little evidence that gender 269 homophily is detrimental to research quality, in that high-impact journals tended to have information is given in S1 Table. 301 Calculating α, the coefficient of homophily  We also calculated α for papers with 2, 3, 4 or ≥5 authors, for all journals that had at least 322 50 suitable papers from 2015-2016 with the specified author list length.

323
Our test assumes that the expected value of α is zero if authors randomly assort, but for an estimate of the extent to which multidisciplinarity within journals inflates α .

352
As well as varying between disciplines, the gender ratio of authors has changed markedly over

374
Calculating standardised journal impact factor 375 We obtained the 3-year impact factor for each journal from Clarivate Analytics (formerly 376 ISI). To account for large differences in impact factor between disciplines, we took the the 377 residuals from a model with log 10 impact factor as the response and the research discipline of  To test whether α for last authors tends to be higher than α for first authors for any given      Positive values mean that α was higher when calculated for first authors, and negative values 630 mean α was higher when calculated for last authors. The mean is very slightly higher than 631 zero, indicating that α tends to be higher for first authors.   Table. Sample sizes for the two datasets, which comprise papers published in the timeframes