Quantifying hierarchy and dynamics in US faculty hiring and retention

The study investigates how faculty hiring and retention shape the composition and dynamics of the US professoriate, with particular attention to inequalities of prestige, production, and gender across fields and domains. Prior work shows that prestige strongly structures hiring networks and outcomes across disciplines, but less is known about how attrition and international training interact with these hierarchies. The authors ask: (1) How unequal is faculty production across institutions and fields? (2) How do attrition patterns interact with production and prestige hierarchies? (3) How have gender demographics among tenure-track faculty changed over 2011–2020, and what processes drive these changes? By assembling a comprehensive, cross-field dataset spanning a decade, the study aims to quantify universal versus field-specific patterns and identify dynamic processes that generate and maintain inequalities, providing evidence to inform policies on hiring, retention, and diversification of the professoriate.

Extensive research documents steep prestige hierarchies in academic hiring networks across many disciplines, measured either through external rankings or inferred from network structure. Prestige confers advantages in review outcomes, resources, productivity, citations, awards, and early-career wages. Prior studies show that most faculty work at less prestigious institutions than their doctoral training institutions, and that hierarchies are relatively stable over time with social closure effects. Evidence also indicates elevated mid-career attrition for women in science and engineering (but not mathematics) and for foreign-born faculty, suggesting that factors beyond prestige structure the professoriate. However, earlier work has often been limited to single fields or static snapshots, with scarce cross-field, longitudinal comparisons and limited treatment of internationally trained faculty. This study builds on that literature by providing a comprehensive, multi-field, decade-long analysis that simultaneously considers hiring, retention, international training, and gender dynamics.

Data: Census-based dataset (via AARC) of tenured or tenure-track faculty employed 2011–2020 at 368 PhD-granting US universities. Inclusion required presence in the majority of sampled years, yielding n=295,089 faculty across 10,612 departments. Faculty records include doctoral university, year of doctorate, rank, and gender. Departments were organized into 107 fields and 8 broad domains; field labels are not mutually exclusive, and analyses used primary appointments for multi-department faculty. Doctoral universities were manually annotated by country. Gender used self-reports when available and otherwise algorithmically inferred (binary man/woman) from name–gender associations; 85% of records had gender annotations. Changes year-to-year were used to annotate new hiring, retention, and attrition events. Networks: For each field, domain, and academia overall, constructed a directed faculty hiring network where nodes are universities and edges u→v denote a doctorate from u leading to a faculty position at v; self-hires are u→u. Aggregations to domains/overall took unions across fields to avoid double-counting multi-field faculty. Analyses: Quantified degree distributions and faculty production inequality using Lorenz curves and Gini coefficients. Compared production vs department size using Kolmogorov–Smirnov tests with Benjamini–Hochberg corrections. Assessed international training composition by continent and specific countries (Canada, UK). Estimated annual per-capita attrition risk ratios for groups by doctoral country: US, Canada/UK, and all other countries; significance via χ² tests with Benjamini–Hochberg corrections. Modeled attrition as a function of the doctoral university’s production rank using logistic regression at field, domain, and overall levels; significance via two-sided t-tests with multiple-comparison correction. Computed Gini coefficients separately for newly hired and existing (sitting) faculty and annually to assess temporal stability. Quantified prestige hierarchies by ranking universities from hiring networks (Methods referenced tools include network ranking models and degree-sequence–preserving null models). Evaluated hierarchy steepness and mobility by comparing empirical upward movement proportions to 1,000 draws from a degree- and production-constrained network rewiring null model. Examined gender dynamics: trends in women’s representation over time, differences between new hires and attritions, and representation by career age (years since doctorate). Assessed self-hiring prevalence and its association with prestige, as well as differential attrition of self-hires. Statistical significance thresholds were α=0.05 with Benjamini–Hochberg corrections where applicable.

• Credentials: 92.7% of US faculty hold doctorates. Non-doctorate prevalence varies widely by domain (e.g., 19% in Humanities vs 1% in Social Sciences), with Theatre, Art History, Music, and English driving much of the Humanities’ non-doctorate share.
• International training: 11% of US faculty have non-US doctorates, varying by domain (2% in Education to 19% in Natural Sciences). Among internationally trained faculty, 35.5% earned doctorates in the UK or Canada, versus 5.4% from all of Africa and the Americas excluding Canada.
• Attrition by doctoral country: Faculty with Canada/UK doctorates (n=11,156) have attrition risks indistinguishable from or slightly lower than US-trained faculty (n=238,676). Faculty with doctorates from other non-US countries (n=20,689) have markedly higher attrition risks overall, across all eight domains, and in 39 of 107 fields; no field shows significantly lower attrition for this group.
• Universal production inequality: Among domestically trained faculty, 80% were trained at just 20.4% of US universities. The top five producers (UC Berkeley, Harvard, Michigan, Wisconsin–Madison, Stanford) account for 13.8% of domestically trained faculty. University faculty size correlates with production (Pearson ρ=0.76, P<1e−5) but does not explain production inequality; production vs size distributions differ in academia overall, all eight domains, and 91/107 fields (K–S tests). Excluding self-hires reduces but does not eliminate these differences in most fields.
• Inequality levels: Overall Gini for domestic faculty production is 0.75, exceeding seven of eight domains. Lowest domain Ginis include Education (0.67) and Medicine and Health (0.67); highest is Humanities (0.77). Domain-level inequality typically exceeds that within constituent fields (Gdomain>Gfield for 104/107 fields).
• Attrition exacerbates inequality: Existing faculty exhibit higher production inequality than newly hired faculty in every field, domain, and overall. Cross-sectional Ginis are stable year-to-year, indicating that post-hire dynamics increase inequality. Logistic models show higher attrition among faculty trained at lower-producing universities in academia overall, all eight domains, and 49/107 fields; those not from top producers are nearly twice as likely to leave annually.
• Gender dynamics: Men constitute 64% of tenure-track faculty. Women’s representation increased from 2011 to 2020 in academia overall, all domains, and 80/107 fields (e.g., Engineering from 12.5% to 17.1%; Education from 55.4% to 58.5%). Nursing is the only field with a significant decrease. However, the share of women among new hires did not increase in 100/107 fields and decreased in 7 fields. New hires are more likely to be women than departing faculty in academia, all domains, and 103/107 fields, a pattern driven by male-skewed retirements; women’s representation by career age shows fewer women among retirement-age cohorts. Despite overall increases, women remain under-represented among new hires in most fields, especially STEM, making future parity unlikely without changes.
• Self-hiring: About 9.1% of all US professors (11% of US-trained) are employed by their doctoral university, with elevated rates at elite institutions. Self-hires leave at higher rates than non-self-hires in most fields, all domains, and overall, indicating that high self-hiring rates persist despite elevated attrition.
• Prestige hierarchies and mobility: The hiring network exhibits steep prestige hierarchies; approximately 71% of faculty move down in prestige from doctorate to job, 18% move up, and 11% are self-hires. Upward movers are few and ascend modestly; downward moves are larger in magnitude. No significant gender differences in mobility exist in 82/107 fields; in 25 fields (many in Medicine and Health), women are less likely to move down and more likely to be self-hires. Across domains, sitting faculty are more likely to be self-hires as prestige increases; new hires are more likely to be internationally trained as prestige increases. Overall, higher-prestige institutions have a greater proportion of men among both new and existing faculty.

The analysis demonstrates a universal core–periphery structure in US faculty hiring across fields and domains: a small core of elite universities trains a large majority of faculty and exports faculty broadly to the periphery, while import flows back to the core or from abroad are limited. Inequalities in faculty production are not explained by department size and are intensified by differential attrition, particularly among faculty trained at lower-producing universities, internationally trained scholars from countries other than the US/Canada/UK, and self-hires. These dynamics produce a form of dynamic equilibrium: inclusivity or diversification gains at hiring are counterbalanced by higher attrition in specific groups, preserving steep hierarchies and unequal representation. Gender results reveal that recent increases in women’s overall representation are largely due to demographic turnover (male-skewed retirements) rather than recent hiring practices, which have remained relatively flat or even trended downward in some fields. Without changes that increase women’s share among new hires, particularly in STEM, progress toward parity is likely to stall. Elevated self-hiring—especially at elite universities—raises concerns about reduced circulation of ideas and potential impacts on scholarship quality, though current high rates appear to be sustained despite higher attrition among self-hires and may reflect older cohorts’ practices. Collectively, these findings highlight how hiring and retention jointly shape academic hierarchies, with prestige amplifying advantages in placement, retention, and demographic composition.

This work provides a comprehensive, cross-field, decade-long quantification of US faculty hiring and retention, revealing universal and steep prestige-driven inequalities in faculty production, elevated attrition among internationally trained scholars from countries other than the US/Canada/UK, higher attrition among faculty trained at lower-producing universities, and persistent gender imbalances driven by cohort turnover rather than recent hiring. The results clarify that attrition dynamics exacerbate inequalities established at hiring, sustaining a core–periphery structure across fields and domains. Future research should (1) causally identify drivers of differential attrition (e.g., institutional support, visa/citizenship status, resources), (2) incorporate doctoral department-level data to refine self-hiring and field-specific production estimates, (3) integrate self-identified, intersectional demographic data to understand representation and attrition across identities, and (4) evaluate interventions aimed at increasing women’s and internationally trained scholars’ retention and improving equity in hiring and mobility across prestige hierarchies.

• Doctoral departments are unknown; self-hiring is measured at the university level and thus overestimates department-level self-hiring. Field-level production and prestige inequalities reflect university placement volumes, not necessarily field-specific PhD graduate production or cross-field hiring flows.
• Demographic data are limited: no self-identified race/ethnicity, socioeconomic status, nationality/citizenship, or place of birth. Gender is binary and often inferred from names, excluding non-binary identities and potentially misclassifying some records. Analyses of international training cannot distinguish foreign-born US-trained faculty from US-born internationally trained faculty.
• The study is observational and does not identify causal mechanisms underlying hiring, mobility, or attrition patterns. Network-based prestige measures and null models, while informative, cannot fully capture unobserved confounders (e.g., subfield demand, institutional policies, resources).
• Aggregation choices (fields, domains) and the necessity to union multi-field appointments may introduce measurement noise. Despite corrections for multiple comparisons, some field-level inferences may be underpowered or sensitive to small sample sizes in niche areas.
• Time frame (2011–2020) may miss earlier or later shifts in hiring practices and demographics; cohort effects indicate non-stationarity over longer horizons.