Quantifying hierarchy and dynamics in US faculty hiring and retention

Faculty hiring and retention are crucial for shaping the US academic workforce, influencing educational outcomes, careers, the spread of ideas, and research priorities. These processes are dynamic, reflecting societal and academic shifts, generational turnover, and efforts to diversify the professoriate by gender, race, and socioeconomic status. Existing research highlights the significant role of prestige in structuring the US professoriate, with prestigious departments disproportionately supplying faculty and exhibiting social closure. Prestige impacts various aspects of academic life, including publication rates, resource access, citations, and awards. However, less is known about the processes of attrition, which, along with hiring, shapes the composition of the faculty. Studies suggest higher attrition rates among women in certain fields and foreign-born faculty. The difficulty of assembling comprehensive data across fields, universities, and time has limited previous analyses. This study aims to address this gap by providing a comprehensive, cross-disciplinary understanding of academic hierarchies and their relationship to persistent social and epistemic inequalities, informing policies aimed at diversifying the professoriate and accelerating scientific discovery.

Previous research extensively demonstrates the influence of prestige on faculty hiring. Studies consistently show that prestigious departments supply a disproportionate number of faculty across various fields, regardless of how prestige is measured. This prestige-driven hiring creates a self-reinforcing cycle, or 'social closure', where prestigious institutions preferentially hire graduates from similarly prestigious institutions. This pattern is supported by both empirical observations and mathematical models of network dynamics. The broad impacts of prestige are well-documented, improving publication acceptance rates, increasing research output and citations, and enhancing career prospects for faculty and graduates. However, less attention has been given to attrition patterns, although evidence suggests higher attrition rates among women in certain STEM fields and foreign-born faculty. These studies highlight a need for a more comprehensive analysis that considers both hiring and attrition processes across diverse fields and universities.

The analysis uses data on tenured or tenure-track faculty employed between 2011 and 2020 at 368 PhD-granting US universities. The dataset includes information on each faculty member's doctoral university, year of doctorate, faculty rank, and gender. After cleaning and preprocessing, the dataset included 295,089 faculty in 10,612 departments, organized into 107 fields and eight domains. Departments were manually annotated by country of doctoral origin. Gender annotations were obtained through self-reporting when available, otherwise using algorithmic annotation based on name-gender associations. Faculty without gender annotations were excluded from gender analyses. Hiring, retention, and attrition were annotated by comparing data from adjacent years. Faculty hiring networks were created for each field, domain, and academia as a whole. In these networks, nodes represent universities, and directed edges represent individuals with doctorates from one university becoming professors at another. Self-hires (faculty employed at their doctoral universities) are represented as self-loops. Aggregate networks for domains and academia were created by taking the union of edges from constituent fields. Anonymized data are publicly available.

The study reveals several key findings: 1. **Universal Production Inequality:** A small minority of universities (20.4%) train a large majority (80%) of domestically trained faculty. The top five universities account for 13.8% of domestically trained faculty. This inequality persists across all domains of study. The Gini coefficient, a measure of inequality, is high (0.75) for the entire academic system, indicating extreme inequality in faculty production. This inequality is rooted in hiring patterns but is exacerbated by attrition. Universities producing fewer faculty exhibit significantly higher attrition rates among their graduates. 2. **Differential Attrition:** Faculty trained outside the US, Canada, and the UK experience markedly higher attrition rates compared to US-trained faculty. Similarly, faculty trained at universities with lower faculty production also have higher attrition rates. This suggests a dynamic equilibrium where the initial inequality in faculty production is further amplified by differential attrition rates. 3. **Gender Inequality:** While women's representation increased in many fields and domains between 2011 and 2020, the gains are primarily due to demographic turnover (higher attrition among older male faculty) rather than increased representation among new hires. The proportion of women among new hires remained relatively flat during this period, suggesting limited progress toward gender parity, particularly in STEM fields. 4. **Self-Hiring:** Approximately 9.1% of US professors are employed by their doctoral universities. Self-hiring rates are surprisingly high across fields and universities, despite negative academic norms. This is more prevalent at elite universities. However, these self-hiring patterns are driven more by previous hiring and attrition patterns rather than recent hiring trends. Self-hires tend to leave academia at higher rates than non-self-hires. 5. **Prestige Hierarchies:** The study shows steep hierarchies of prestige in faculty hiring. Prestige significantly influences faculty mobility, with a tendency for faculty to move to universities of similar or higher prestige, creating a core-periphery structure in the academic system. Prestige is linked to various outcomes, including self-hiring rates and the likelihood of hiring internationally trained faculty. However, the relationship between prestige and gender is complex and varies across domains.

The findings reveal a system-wide pattern of inequality in the US academic workforce shaped by the interplay of hiring and attrition processes. A small number of elite universities dominate faculty production, which is then amplified by differential attrition rates affecting underrepresented groups. Gains in gender representation are largely a consequence of demographic shifts rather than fundamental changes in hiring practices. High self-hiring rates at elite institutions present a paradox, contradicting established norms and potentially hindering the spread of ideas. The study highlights the complex relationship between prestige, hiring patterns, and attrition, underscoring the need for interventions focused on both improving hiring practices and enhancing retention strategies for underrepresented groups. The core-periphery structure of the academic system necessitates targeted strategies to promote broader participation and equity.

This study provides a comprehensive quantitative analysis of US faculty hiring and retention, revealing widespread inequalities related to university prestige, geographical origin of doctoral training, and gender. The findings highlight the complex interplay between hiring patterns and attrition, emphasizing the need for targeted interventions to mitigate these inequalities and foster a more diverse and equitable academic workforce. Future research should focus on the underlying mechanisms driving differential attrition rates and explore the long-term implications of the observed patterns for scientific discovery and knowledge production.

While the study uses a large and comprehensive dataset, several limitations exist. The data lack information on doctoral departments, leading to estimates of self-hiring that may be upper bounds. The analysis considers country of doctoral training but not citizenship or birth, limiting inferences about foreign-born, US-trained faculty. The reliance on name-gender associations for gender annotation limits the study's ability to consider more expansive gender identities. The study reveals correlations but doesn't establish causal relationships between observed patterns and underlying mechanisms. Future research should address these limitations.