Quantifying Attrition in Science: A Cohort-Based, Longitudinal Study of Scientists in 38 OECD Countries

Abstract: In this paper, we explore how members of the scientific community leave academic science and how attrition (defined as ceasing to publish) differs across genders, academic disciplines, and over time. Our approach is cohort-based and longitudinal: We track individual male and female scientists over time and quantify the phenomenon traditionally referred to as 'leaving science.' Using publication metadata from Scopus - a global bibliometric database of publications and citations - we follow the details of the publishing careers of scientists from 38 OECD countries who started publishing in 2000 (N = 142,776) and 2010 (N = 232,843). Our study is restricted to 16 STEMM disciplines (science, technology, engineering, mathematics, and medicine), and we track the individual scholarly output of the two cohorts until 2022. We use survival analysis to compare attrition of men and women scientists. With more women in science and more women within cohorts, attrition is becoming ever less gendered. In addition to the combined aggregated changes at the level of all STEMM disciplines, widely nuanced changes were found to occur at the discipline level and over time. Attrition in science means different things for men versus women depending on the discipline; moreover, it means different things for scientists from different cohorts entering the scientific workforce. Finally, global bibliometric datasets were tested in the current study, opening new opportunities to explore gender and disciplinary differences in attrition.

• The paper quantifies scientist attrition using cohort survival analysis to reveal evolving gender and disciplinary trends.
• The paper finds that early cohorts showed higher attrition among women, but gender differences are diminishing in more recent cohorts.
• The paper identifies disciplinary variations, with math-intensive fields exhibiting unique attrition patterns compared to other scientific domains.

Gender and Disciplinary Differences in Scientist Attrition Across OECD Countries: A Longitudinal Analysis

The paper by Kwiek and Szymula offers a comprehensive and nuanced quantitative assessment of scientist attrition spanning 38 OECD countries. The study rigorously investigates how gender and disciplinary characteristics influence scientists’ transitions out of academia, articulated through the lens of publishing activity as captured by the Scopus database. Deploying survival analysis methodologies, it dissects attrition trajectories for two large cohorts of scientists who commenced publishing in 2000 and 2010, respectively, and followed their careers until 2022.

Key Findings

The study highlights critical insights into how attrition patterns have evolved across genders and disciplines. Traditionally, women have been more likely than men to exit scientific publishing, a trend explained by various systemic factors, including unfavorable workplace climates and gender-based barriers. However, the findings indicate this gender differential may be shrinking. Notably, attrition became less gendered for the 2010 cohort compared to the 2000 cohort.

• Gender Impact on Attrition: For the 2000 cohort, women exhibited a significantly higher attrition rate than men, with approximately one-third leaving after five years and about two-thirds ceasing to publish by 2019. In contrast, the 2010 cohort displayed minimal gender differences, suggesting a potential generational shift in gender dynamics within scientific careers.
• Disciplinary Variations: The attrition effect is not uniform across disciplines. The research highlights that in highly math-intensive domains, such as physics and astronomy (PHYS), and computer science (COMP), female attrition rates did not significantly differ from male rates, even though these fields have historically lower female representation. This indicates that women in these sectors may exhibit particular resilience or experience different challenges that do not translate into higher attrition when measured by publishing activity.
• Cohort Comparison: The survival and hazard rate analyses delineate a steeper decline in publishing persistence for later cohorts. By 2022, over half of the scientists from the 2010 cohort had ceased publishing within a mere nine years, regardless of gender—a different pattern from that observed in the 2000 cohort, where gender differences were more pronounced over time.

Implications and Future Directions

This study bears significant implications for policy development and institutional practices within higher education and research institutions. The apparent diminution of gender disparity in attrition among more recent cohorts may reflect evolving institutional cultures and broader systemic changes, such as increased efforts toward gender equity or shifts in support structures that aid retention of women in traditionally male-dominated fields. However, the persistent high attrition rates across both genders for the 2010 cohort emphasize an ongoing challenge to retain talent in the scientific workforce.

Further research could benefit from exploring the detailed interactions between institutional policies, such as parental leave or mentorship programs, and their role in shaping attrition trends, as well as leveraging administrative data in combination with bibliometrics to gain a deeper understanding of career trajectories. Additionally, exploring the dynamics in non-OECD countries and in non-STEMM disciplines would provide a more global perspective on scientist retention trends. These insights are essential for crafting interventions that could mitigate early career attrition and foster a more inclusive and sustainable scientific environment.

Overall, this study contributes valuable empirical evidence and methodological innovation to the field of the science of science, suggesting that while gender disparities in academic science are shrinking, attritional challenges persist, warranting continued scholarly and policy attention.