Gender Bias Impacts Top-Merited Candidates

Abstract

Expectations of fair competition underlie the assumption that academia is a meritocracy. However, bias may reinforce gender inequality in peer review processes, unfairly eliminating outstanding individuals. Here, we ask whether applicant gender biases peer review in a country top ranked for gender equality. We analyzed peer review assessments for recruitment grants at a Swedish medical university, Karolinska Institutet (KI), during four consecutive years (2014-2017) for Assistant Professor (n = 207) and Senior Researcher (n = 153). We derived a composite bibliometric score to quantify applicant productivity and compared this score with subjective external (non-KI) peer reviewer scores of applicants' merits to test their association for men and women, separately. To determine whether there was gender segregation in research fields, we analyzed publication list MeSH terms, for men and women, and analyzed their overlap. There was no gendered MeSH topic segregation, yet men and women with equal merits are scored unequally by reviewers. Men receive external reviewer scores resulting in stronger associations (steeper slopes) between computed productivity and subjective external reviewer scores, meaning that peer reviewers "reward" men's productivity with proportional merit scores. However, women applying for assistant professor or senior researcher receive only 32 or 92% of the score men receive, respectively, for each additional composite bibliometric score point. As productivity increases, the differences in merit scores between men and women increases. Accumulating gender bias is thus quantifiable and impacts the highest tier of competition, the pool from which successful candidates are ultimately chosen. Track record can be computed, and granting organizations could therefore implement a computed track record as quality control to assess whether bias affects reviewer assessments.

Keywords: bibliometry; diversity; equality; faculty positions; gender; life science; peer review.

Figure 1

Number of applicants and proportions at each step of the recruitment process for the career ladder positions at Karolinska Institutet, divided by each year [(A) 2014, (B) 2015, (C) 2016, (D) 2017] for applications for assistant professor positions (A–D) or for Senior Researcher positions (E–H). In five of the eight calls, the % of men awarded grants was higher than the % of men in the eligible applicant pool (A, D–G).

Figure 2

Analysis of field representation among men and women applicants using MeSH (medical subject headings) term analysis. (A) Applicants for assistant professor positions are more diverse and encompass 207 applicants (74 women and 133 men). Field representation is not obviously skewed by biased fields such as cardiovascular medicine, which is represented by MeSH terms among both men and women. There is a high degree of overlap in MeSH terms among men and women applicants as seen in the Venn diagram at right. (B) Men and women applicants for Senior Researcher positions were fewer in total (153), are highly similar, and show near-complete overlap in MeSH terms, as seen in the Venn diagram at the far right. (C) Analysis of MeSH terms for gender-segregated topics revealed a very small fraction of gender-segregated MeSH terms among men (1.2%) or women (1.3%) applicants.

Figure 3

Linear regression associations between composite bibliometric scores and external reviewer scores received on merits combined for all years (2014–2017) and stratified by gender for applications for (A) Assistant Professor and (B) Senior Researcher positions.

Figure 4

Proposed three-step review process to minimize impact of bias. Project plans can be reviewed blinded by reviewers when applicants are junior, reducing risk of bias. A composite bibliometric score is calculated to support assessment of past productivity and impact, reducing bias (some metrics have been shown to be unfairly biased by gender). An individual assessment, performed by a peer reviewer, integrates scores from the project plan and the composite bibliometric score, to assess the feasibility of the project. Together, these three steps reduce bias while allowing for an assessment of a project's innovativeness and an applicant's competence to execute the project.