ORCID Number

0000-0002-8713-6300

Date of Award

Summer 2026

Access Type

Dissertation - Open Access

Degree Name

Doctor of Philosophy in Aviation

Department

College of Aviation

Committee Chair

Dothang Truong

Committee Chair Email

truongd@erau.edu

First Committee Member

Frank Ayers

First Committee Member Email

ayersf@erau.edu

Second Committee Member

Michael McCormick

Second Committee Member Email

mccormm9@erau.edu

Third Committee Member

Arun P. Saini

Third Committee Member Email

arun.saini@gmail.com

College Dean

Alan J. Stolzer

Abstract

Air traffic control (ATC) tower safety efficiency remains inadequately measured despite critical workforce challenges, including staffing shortages, training pipeline constraints, and geographic retention difficulties. This mixed-methods study developed and validated a comprehensive framework for measuring tower safety efficiency by integrating subject-matter expert insights with multi-model data envelopment analysis (DEA) across 66 Federal Aviation Administration (FAA) tower facilities from 2016 to 2024. Six focus group interviews with experienced professionals (100% with > 20 years of experience), averaging 96.4 minutes, identified critical efficiency factors through systematic coding, achieving substantial inter-rater reliability (Cohen's kappa 𝜅 = .68). Qualitative findings showed that staffing shortages don't just increase workload—they trigger a cascading breakdown across training quality, duty hours, and rest adequacy that ultimately compromises tower safety, a relationship all six subject matter experts confirmed. A single traffic complexity factor comprised of mixed aircraft types was also identified as a powerful safety efficiency driver.

The quantitative analysis examined 22 variables across human resources, environmental factors, operational complexity, and safety performance indicators. A Friedman test showed model-dependent patterns: super-efficiency (p = .532) and CCR (p = .499) models showed non-significant variation, while the SBM model (p < .001) and two-stage network models revealed significant changes (p = .002), confirmed by Bonferroni-corrected pairwise comparisons. The BCC model showed a significant overall difference (p < .001) but no significant corrected pairwise differences.

Bootstrapping validation (1,000 iterations) of the deterministic 2024 CCR results revealed 21 Stage I towers exhibited low bias levels (< .05) with three of those displaying narrow confidence intervals (≤ 0.10). Additionally, 41 Stage II towers exhibited a low bias level along with 17 of those having narrow confidence intervals. A complementary leave-one-out sensitivity analysis found extreme rank instability in Stage I (rank ranges up to 60 of 65 positions) but substantially greater stability in Stage II. A cross-stage classification panel identified four consistently efficient towers (CMA, FFZ, FRG, and SAN) and 18 towers that were ever inefficient (SNA, CNO, BOS, FXE, OAK, MIA, FLL, IAD, PDX, LGB, TMB, BUR, SEE, AUS, BWI, FTW, LGA, and RDU).

While SBM-DEA proved statistically superior at detecting genuine operational fluctuations, the coupled bootstrap and leave-one-out diagnostics reveal that standard deterministic Stage I rankings are fundamentally unreliable for managerial decision-making without explicit uncertainty assessment. It may be, therefore, necessary to establish bootstrapping and DMU-removal sensitivity analysis as necessary safeguards, not optional supplements, for defensible DEA application in aviation safety benchmarking.

This framework offers two key quantitative contributions: a two-stage decomposition separating operational capacity from safety efficiency, and a multi-model DEA triangulation tool distinguishing genuine performance from measurement artifacts. Practical applications include guiding investigation at validated inefficient facilities and informing priority-setting for training pipelines, locality pay structures, and airspace complexity.

Additional Files
Gilbert GS9_Acceptance.pdf (395 kB)
Download

Share

COinS