Diversity trends among faculty in STEM and non-STEM fields at selective public universities in the U.S. from 2016 to 2023

The study examines how faculty diversity has evolved in STEM versus non-STEM fields at selective U.S. public universities since the 2015–16 campus protests and subsequent institutional diversity, equity, and inclusion (DEI) investments. The research question centers on whether these post-2016 diversity efforts, alongside broader dynamics, differentially influenced racial-ethnic and gender representation across fields and ranks. Contextual motivation includes longstanding claims that faculty diversity enhances curricular breadth, student belonging, and counters stereotypes, and evidence that demographic match between students and instructors affects student outcomes and major choice. The authors aim to provide field-specific trend evidence that large-scale datasets typically cannot offer due to lack of field identifiers, thereby informing equity implications within academia and downstream labor market outcomes given the higher returns to STEM degrees.

Prior work documents underrepresentation of Black, Hispanic, and female faculty—especially in STEM—and links part of the STEM/non-STEM imbalance to the PhD pipeline (Li & Koedel, 2017; Ginther et al., 2010; Parsons, 2023). Universities scaled up DEI initiatives post-2015–16 protests, including major funding commitments (e.g., Michigan, Yale), establishment of Chief Diversity Officers, diversity advisors on hiring committees, and diversity-augmenting postdoctoral programs. Job postings increasingly mention or require diversity statements. Li and Koedel (2017) warned untargeted diversity efforts could exacerbate field imbalances because underrepresented candidates are more prevalent in non-STEM pipelines. A substantial literature shows student-instructor demographic match improves course completion, persistence, and major choice (Bettinger & Long, 2005; Carrell et al., 2010; Fairlie et al., 2014; Price, 2010; among others), implying faculty diversity distribution across fields can shape student trajectories and future faculty pipelines.

Data sources: (1) IPEDS for public R1 universities, covering academic years 2001–02 to 2021–22, with 102,592–115,297 faculty annually. IPEDS offers comprehensive demographic trends but lacks faculty field identifiers. (2) A panel extension of the Li & Koedel (2017) dataset (L&K), which identifies faculty by field. L&K covers all tenured and tenure-track faculty in 120 departments across 40 selective public universities (approximately top-40 publics by 2016 U.S. News). Wave 1 (2015–16) includes 4,139 faculty; Wave 2 (2022–23) includes 3,851. Departments: biology, chemistry, economics (treated as STEM), and English, educational leadership and policy, sociology (non-STEM). A sensitivity analysis uses a stricter definition (STEM: biology, chemistry; non-STEM: English, sociology) with substantively similar results. Coding: In L&K, race-ethnicity and sex were externally coded using faculty photos, names, and biographical details; interrater reliability was high (2015–16: 95.5% race-ethnicity, 99.75% sex; 2022–23: 96% and 100%). IPEDS reports gender (historically ambiguous relative to sex; revisions begin 2023–24). Baseline comparability: 2016 summary statistics show the 40-university IPEDS subsample resembles all public R1s; IPEDS and L&K are broadly aligned, with small differences due to category coverage (e.g., multiracial, nonresident aliens, and unknown race in IPEDS) and L&K’s six-field scope. Analytic approach: Descriptive trends from IPEDS (overall and by rank) for the L&K university sample from 2002–2022; field-specific trends from L&K (2016–2023). To assess post-2016 changes, interrupted time series (ITS) regressions estimate deviations from pre-protest linear trends: Y_it = β0 + β1 T_it + Σ_x I_xit β_x + ε_it, where β_x captures deviations for years 2017–2022. Focus is on assistant professors given hiring-centric initiatives and typical entry at this rank. Regressions are weighted by university-year faculty counts; standard errors clustered by university. Sensitivity: Trends are similar when using all public R1s in IPEDS; results hold under stricter STEM/non-STEM definitions (some loss of significance for Black assistant professors). Data availability: Harvard Dataverse (doi:10.7910/DVN/XK70EI).

Baseline (2016, L&K): Large cross-field disparities: Black faculty share 1.3% in STEM vs 9.9% in non-STEM; Hispanic 3.3% vs 5.2%; Asian 15.0% vs 7.0%; White 80.3% vs 76.9%. Women comprise 25.5% in STEM vs 48.6% in non-STEM. Long-run trends (2002–2022, IPEDS): White faculty declined from 83% to 66%. Asian increased from 7.7% to 16.0%. Black rose slightly from 3.3% to 3.8%, with a sharp increase in the Black assistant professor share after 2016 (nearly 40% growth from a 4.2% base). Hispanic roughly doubled from slightly below the Black share in 2002. Women increased from 25% to 36% overall; women are 47% of assistant professors by 2022, and the female share among full professors rose by 13 percentage points (2002–2022), narrowing rank gaps. Post-2016 deviations (ITS, assistant professors): Black assistant professor share deviates positively from pre-trend, with significant increases in 2021 and 2022 (p≤0.10). White assistant professors show an increasingly negative deviation, significant in 2022 (p≤0.10). Asian assistant professor deviations are negative post-2016 but reflect a flattening that began pre-2016; Hispanic and female assistant professor deviations are not statistically significant. Field heterogeneity since 2016 (L&K 2016–2023): Assistant professor changes are largest and statistically significant across fields. Black and Hispanic assistant professor shares increased more in non-STEM than STEM, widening racial-ethnic gaps. Female assistant professor share increased in STEM and declined slightly in non-STEM, narrowing the cross-field gender gap. Across all ranks, differences are small and generally not significant; for all faculty, female share growth is about 1.3 percentage points higher in STEM (4.5 vs 3.2), and Black share growth is about 0.8 percentage points higher in non-STEM (1.2 vs 0.4), but these overall differences are not statistically significant.

The findings indicate heterogeneous post-2016 diversity trajectories: racial-ethnic gaps between STEM and non-STEM fields are widening (especially for Black faculty), while gender gaps are narrowing due to faster gains for women in STEM. Concentration of changes among assistant professors aligns with hiring-centric DEI initiatives that would affect inflows more than the stock of senior faculty, though broader pre-existing dynamics also contribute (e.g., accelerated declines in White full professors absent explicit policies). Given strong evidence that demographic match with instructors influences course-taking, major persistence, and attainment, widening racial-ethnic disparities across fields may further reduce Black and Hispanic representation in STEM, with implications for both student equity and future faculty pipelines. Conversely, gender trends suggest progress toward narrowing cross-field gender disparities in STEM, potentially reflecting longstanding emphasis on gender diversity in STEM (e.g., NSF ADVANCE). A key policy implication is that institutions should monitor and manage cross-field balance in diversity gains; untargeted efforts may unintentionally reinforce existing disparities where pipelines are already uneven by field.

This study contributes a field-specific panel analysis of faculty diversity trends at selective public universities from 2016 to 2023, combining comprehensive IPEDS trends with a unique department-level dataset identifying STEM versus non-STEM fields. The principal contribution is documenting that, post-2016, female representation has risen faster in STEM—helping to narrow gender gaps—while Black and Hispanic representation has grown more in non-STEM—widening racial-ethnic gaps across fields. These effects are concentrated at the assistant professor level, consistent with hiring-focused initiatives. Policy recommendations include explicitly accounting for field imbalances when designing diversity efforts to avoid unintended consequences for student representation and future faculty pipelines. Future research should: (1) extend the time horizon to assess how entry-cohort changes translate into stock changes across ranks, (2) use larger samples to examine intersectional race-gender trends by field, and (3) relate faculty diversity trends to student diversity and outcomes by field while addressing endogeneity.

- Field coverage: The L&K dataset includes only six departments across 40 selective public universities; results may not generalize to all fields or institutions. - Measurement: L&K race-ethnicity and sex coding relies on external assessment, though interrater reliability is high; IPEDS’ gender/sex categorization has been historically ambiguous. - Granularity: Sample sizes are insufficient to analyze intersectional groups (e.g., Black women) by field and rank. - Causality: Interrupted time series captures deviations from pre-2016 trends but cannot isolate causal effects of specific university policies; other contemporaneous dynamics likely contribute. - Data timing: IPEDS lacks 2023 data and provides no field identifiers; optional demographic reporting in certain years prior to 2017 limits continuous trend coverage.