Trends in Scientific Collaboration Between Men and Women Following the #MeToo Movement

Abstract

Background

The #MeToo movement raised global awareness about harassment in the workplace. Concerns were raised, however, that the movement may have unintendedly harmed women in academia by decreasing collaboration invitations from men in senior positions, who might be more reluctant to collaborate.

Objective

To analyze whether collaborations between first author women and last author men decreased after the #MeToo movement.

Design

Retrospective observational study.

Participants

Names of first and last authors of 122,450 US review articles from the PubMed database published between 2014 and 2020.

Main Measures

Change in the proportion of review articles with a first author woman and a last author man following the peak of the #MeToo movement in October 2017. Additionally, among review articles with a last author man, trends of women first authorship in the USA and Europe (control group) were compared.

Key Results

We analyzed 122,450 review articles with first and last authors from US institutions. Of 85,015 articles by a man last author, 37.5% (31,902) had a woman first author. In contrast, when the last author was a woman, the first author was also a woman in 53.6% of articles (20,078) (p<0.001 for difference). Among review articles with a last author man, there was no change in the proportion of articles with a woman first author before versus after the peak of the #MeToo movement (e.g., p=0.13 for difference between the 12 months following October 2017 compared to the pre-#Me-too period). Among European institutions, of 72,036 articles by a man last author, 43.4% (31,294) had a woman first author, higher than the proportion observed in the USA. Trends in collaboration between first author women and last author men were similar in the USA and Europe after the peak of the #MeToo movement (p=0.65).

Conclusions

The #MeToo movement was not associated with a reduction in the rate of scientific review article authorship collaborations between first author women and last author men in the life sciences. These findings, if generalizable, suggest it is possible to promote accountability for harassment in the workplace without limiting decreases in collaboration.

INTRODUCTION

Following several reports of sexual harassment by high-profile men in the Fall of 2017, the #MeToo movement raised global awareness about harassment in the workplace, including in medicine and the life sciences,^1–5 where sexual harassment and gender discrimination in the workplace and in educational settings have been widely documented.^6–8 For example, in a large survey of academic physicians, one-third of women reported experiencing sexual harassment compared with 4% of men surveyed.⁸ In this setting, the #MeToo movement provided a critical opportunity to increase awareness of sexual harassment and gender discrimination in medicine and to enact systemic change.

Concerns were raised, however, primarily by men, that increased attention to harassment and discrimination in the workplace, while appropriate and needed, may have unintended consequences for mentorship relationships between men and women, including in the field of medicine.^9–11 For example, outside of medicine, one survey found that men reported being more reluctant to have individual meetings with women, particularly those in more junior roles and those that they did not know well.¹² A study of new collaborations between junior women and senior men in US economics departments found decreased collaborations after the #MeToo movement, with a larger reduction in collaborations in those universities where sexual harassment policies were more ambiguous, and perceived risk of sexual harassment accusations for men was higher.¹³

Despite these concerns and the possibility that the #MeToo movement may have also improved mentorship opportunities, by raising awareness of systematic discrimination against women in the workplace, no large-scale evidence exists on trends in collaborations between men and women in the life sciences following the #MeToo movement. We therefore analyzed the gender compositions of authors of more than 100,000 PubMed-indexed review articles in the life sciences in the years surrounding #MeToo. Our objective was to determine whether research partnerships between first author women and last author men decreased after the peak of the #MeToo movement.

METHODS

Overview

In this retrospective, observational study, we compared gender compositions of first and last authors for life sciences review articles indexed in the PubMed database, which reflects the most comprehensive reference list of scientific articles published in the life sciences. We focused on first and last authorship since in the life sciences these positions often, though not always, correspond to more junior and senior authorship roles. Articles with a single author were excluded from this analysis. Analyses were restricted to scientific review articles, for two main reasons. First, in review articles the invitation to collaborate may often come from a senior author, allowing us to obtain insights about how senior men in academia may have made decisions about scientific collaborations.¹⁴ Second, the publication process is quicker for review articles than for original investigations, enabling us to understand the early potential impact of the #MeToo movement on research collaborations.

Articles published between January 2014 and April 2020 were included in this analysis. Articles published after April 2020 were excluded because of the additional negative impact of the COVID-19 pandemic on research collaborations involving women.¹⁵ We performed analyses of authors in the USA, since the movement was most influential there, and examined how US trends in collaboration between men and women before and after the #MeToo movement compared with trends in Europe, which served as a control group to account for secular trends in authorship that may have, by chance, coincided with the start of the movement.

Data Sources and Definitions

We applied the PubMed advanced search function to filter articles by date range, publication type, and author affiliation, and downloaded metadata for every matched article. We defined the article as American if the PubMed affiliation field identified any American state, and European if it identified any European country. Articles with both American and European affiliations were excluded. For analyses comparing periods pre- and post-#MeToo, the latter was determined as beginning in October 2017, when the #MeToo movement achieved peak awareness, as defined by number of Google searches.¹⁶

Inferred author gender was determined by Genderize.io, a validated algorithm that predicts the probability of being a man or woman from a person’s first name.¹⁷ This algorithm has been used in prior publications.^{15, 18, 19} Importantly, however, gender determination was inferred from names and does not directly represent authors’ sex at birth or self-identified gender. The algorithm may also be less accurate if there are few recorded instances of a given name in the database. To avoid this latter issue, when the database had fewer than 40 people with any given name, the respective article was excluded from analyses. If gender was unidentifiable from the PubMed metadata dataset, due to cases involving missing values or single-letter names, the article was also removed from analyses.

Outcomes

The primary outcome was the change in the proportion of review articles with a first author woman and a last author man following the peak of the #MeToo movement. Moreover, we hypothesized that because the movement was heavily focused in the USA, at least initially, any gender differences in mentorship would be detectable earlier in the USA than in Europe. Thus, in addition, we assessed changes in the proportion of review articles with a woman (vs. man) first author and a last author man after peak #MeToo in the USA vs. Europe.

Statistical Analysis

We began by graphing trends in scientific collaboration between men and women in the USA before and after the #MeToo movement achieved peak awareness (first and last authorship combinations of men and women). For review articles by last author men, we then used linear regression to estimate changes in the monthly percentage of articles authored with a woman first author before and after the peak of the #MeToo movement. The model included an interaction between the month an article was published (indicator variable for month) and an indicator variable for the post-#MeToo period—a statistically significant interaction term would indicate that the trend of authorship between last author men and first author women (vs. men) changed after the movement. Because of the time elapsed between initial invitation to submit review articles and final publishing, it is unclear how long it would take for any potential #MeToo effect to become apparent in review articles. Thus, we separately analyzed the post-#MeToo indicator at varying time points: 6, 12, 18, and 24 months after October 2017.

To assess whether authorship patterns observed in the USA might be explained by general authorship patterns occurring globally, we then conducted an analysis comparing US trends with European trends. Among review articles with last author men, we began by graphing trends in the proportion including first author women for the USA and Europe. We then conducted a difference-in-difference analysis, with Europe as a comparator, to detect diverging authorship trends. Specifically, we estimated a linear regression of the proportion of men last author articles with a woman first author as a function of a post-#MeToo indicator, indicator for the authorship team being in Europe, and a difference-in-difference interaction term between the two variables.

Statistical significance was defined as p<0.05; to limit type I error, interaction p-values were Bonferroni-adjusted. All analyses were conducted in R software (version 3.6.1).²⁰

RESULTS

A total of 122,450 review articles involving first and last authors at American institutions with ascribed genders were analyzed between January 2014 and April 2020 (Table 1). Of 85,015 articles by last author men, 37.5% (31,902) had a women as first author. In contrast, when the last author was a woman, the first author was also a woman in 53.6% (20,078) of articles (chi-square test for difference, p<0.001).

Table 1.

Characteristics of Articles

	USA (N = 122,450)	Europe (N = 105,430)
First author woman, % (N)	42.4 (51,980)	47.2 (49,796)
First author man, % (N)	57.6 (70,470)	52.8 (55,634)
Last author woman, % (N)	30.6 (37,435)	31.7 (33,394)
Last author man, % (N)	69.4 (85,015)	68.3 (72,036)
Articles published by year, N
2014	13,716	10,701
2015	15,651	12,276
2016	16,915	13,535
2017	17,819	13,985
2018	22,093	20,330
2019	25,449	22,690
2020*	10,807	11,913

Distribution of gender in first and last author positions, among review articles with at least 2 authors between 2014 and 2020. Articles in the US column had author affiliations only in the USA and vice versa for articles in the Europe column, i.e., the columns are mutually exclusive *The year 2020 included articles between January and April

Gender trends in collaboration suggested that among US-based review articles with a last author man, there was no change in the proportion of articles with a first author woman before versus after the start of the #MeToo movement (Fig. 1). Specifically, among articles by last author men, there was no statistically significant interaction between month of publication and the #MeToo period in any of the pre-defined time points after adjustment for multiplicity (critical p-value=0.0125): 6 months (p-value for interaction=0.21), 12 months (p=0.13), 18 months (p=0.021), or 24 months (p=0.38) after October 2017.

Figure 1.

Trends in scientific collaboration between men and women following the #MeToo movement, USA. Monthly percentage of each gender combination for first and last authors of review articles between 2014 and 2020. The denominator is the total number of review articles with identifiable genders published by authors in the USA during a given month.

We analyzed 105,430 review articles with first and last authors from European institutions. Of 72,036 articles by last author men, 43.4% (31,294) had a first author woman, higher than the proportion observed in the USA. Overall, of 33,394 articles by last author women, 55.4% (18,502) had a first author woman, similar to the proportion observed in the USA. Following peak #MeToo, authorship trends in the USA and in Europe were comparable (p-value for difference-in-difference estimate = 0.65) (Fig. 2).

Figure 2.

Trends in scientific collaboration between men and women following the #MeToo movement, USA vs. Europe. Monthly percentage of review articles with a woman first author in the USA and Europe between 2014 and 2020. The denominator is the total number of published review articles with identifiable genders and a man last author located in the USA or Europe, respectively, during a given month.

DISCUSSION

Using a large, publicly available database of biomedical research articles indexed in PubMed, we found that the #MeToo movement was not associated with changes in authorship rates between first author women and last author men in the setting of American scientific review articles. Despite early concerns about potential unintended consequences of the movement on mentorship of women in academia by men, particularly senior men, these findings suggest that, on average, mentorship relationships for the specific type of collaboration studied here may not have been negatively impacted. Importantly, these findings reflect average trends that may mask important heterogeneity in the impact of the #MeToo movement. For example, while we did not observe a reduction, on average, in the proportion of articles authored by last author men and first author women, it is possible that the movement may have impacted last author men in opposing ways that, on average, create no observed change (e.g., a subset of last author men may have promoted greater collaborations with first author women, as a result of greater awareness of systemic gender discrimination brought about by #MeToo, while a second subset of may have reduced collaboration invitations).

These findings differ from a related study of new collaborations between junior women and senior men in US economics departments, which found decreased collaborations after the #MeToo movement.¹³ Although the reasons for this difference are unclear, they may stem from high levels of reported sexual harassment and gender discrimination within the field of economics.²¹ For example, in a survey of more than 9000 current and past members of the American Economic Association, 48% of women reported discrimination based on sex, compared with 3% of men.²² An alternative explanation for the difference is the type of collaboration being analyzed. While our study focused on review articles to obtain early insights on mentoring decisions by senior men, it is possible that invitations to collaborate would be most affected for work that is perceived as harder, more time-consuming, and requiring more personal interaction, such as original investigations.

Our study also found that among women last authors, the share of authored review articles involving first author women was similar to the share of articles involving first author men, whereas among last author men the share of articles with first author men substantially exceeded the share of articles with a first author woman, a finding that has been observed elsewhere.^23–25 It is concerning that over the period analyzed, including the period following the #MeToo movement, the proportion of men last author articles that involved a first author woman did not increase and remained below the proportion for women last authors.²⁶ This less frequent collaboration between last author men and first author women remains an unresolved issue with opportunity for improvement.

Limitations

Our study had several limitations. First, we focused on one measure of mentorship and collaboration: co-authorship of scientific review articles. Mentorship and collaborations of other forms—e.g., career mentorship or other research collaborations—may have been affected differently. Relatedly, whether the nature of mentoring within these collaborations or in other scientific settings has changed cannot be addressed by our study. Second, our study design focused on the role of the last author alone in initiating authorship invitations. In reality, last authors may often be part of a larger, multiple primary investigator team, and the decision to invite authorship may be a joint decision by the group. In this case, the potential for reduced invitations to first author women may be mitigated by this group dynamic. Relatedly, although it is common in the life sciences for the first author to be more junior and the last author to be more senior, this is not always the case. Our data did not include information on faculty rank or actual seniority. Our results should be interpreted as studying changes in first–last author collaboration patterns after the peak of the #MeToo movement. Third, our study design did not permit us to identify who was invited to be an author on any given article, only the authorship that resulted from invitations. Fourth, author names that were missing from PubMed metadata or were particularly uncommon led some articles to be excluded from analyses, though this exclusion should not be systematic and therefore would not be expected to affect conclusions. Fifth, the classifications “men” and “women” were inferred from names traditionally associated with a specific gender, and may not represent sex at birth or self-identified gender.

CONCLUSIONS

In conclusion, despite widespread concerns that the #MeToo movement could alter research collaborations between junior women and more senior men in academic medicine, our findings suggest the movement did not prominently affect collaborations between these two groups.

Data Availability

Data is available from Dr. Zimerman by request.

Declarations

Conflict of Interest

Dr. Jena reports receiving (in the last 36 months) consulting fees unrelated to this work from Bioverativ, Merck/Sharp/Dohme, Janssen, Edwards Life Sciences, Amgen, Eisai, Otsuka Pharmaceuticals, Vertex Pharmaceuticals, Sage Therapeutics, Precision Health Economics, and Analysis Group. Dr. Jena also reports receiving (in the last 36 months) income unrelated to this work from hosting the podcast Freakonomics, M.D., from book rights to Doubleday Books, and from speaking fees from AAE and the Harry Walker Agency; Dr. Rotenstein reports research support from the American Medical Association and FeelBetter Inc. The research conducted was independent of any involvement from the sponsor of the study. The study sponsor was not involved in study design, data interpretation, writing, or the decision to submit the article for publication. The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.