ORCID Number

0009-0000-9974-5309

Date of Award

Fall 10-30-2026

Access Type

Thesis - Open Access

Degree Name

Doctor of Philosophy in Aviation Business Administration

Department

Aviation Business Administration

Committee Chair

Ahmed F. Abdelghany

Committee Chair Email

[email protected]

First Committee Member

Vitaly Guzhva

First Committee Member Email

[email protected]

Second Committee Member

Chunyan Yu

Second Committee Member Email

[email protected]

Third Committee Member

Daniel Friedenzohn

Third Committee Member Email

[email protected]

College Dean

Joseph Gibney

Abstract

This study evaluates how the treatment of seat capacity and airfares as endogenous versus exogenous variables influences time-series forecasting accuracy for airport-level origin–destination (O–D) passenger demand. Accurate O–D forecasts are central to airline network planning, revenue management, and capacity allocation, particularly in operating environments characterized by demand volatility and binding supply-side constraints. Although airfares and seat capacity are core airline decision variables, limited large-scale evidence establishes how alternative modeling assumptions about these variables affect forecasting performance across heterogeneous O–D markets.

Monthly data from January 2022 to April 2024 were compiled for the top 2,000 O–D pairs using Sabre Market Intelligence, including passenger demand, average fare, and average daily nonstop seat capacity. Three forecasting specifications were compared. First, a univariate Seasonal Autoregressive Integrated Moving Average (SARIMA) model provided a demand-only benchmark. Second, a Vector Autoregression (VAR) model treated passenger demand, seat capacity, and fare as jointly endogenous, allowing for dynamic feedback among variables. Third, a Seasonal ARIMA with Exogenous Regressors (SARIMAX) model incorporated seat capacity and fare as external predictors. Models were trained using observations through December 2023 and evaluated via out-of-sample forecasts for January to April 2024. Forecast accuracy was assessed using Standardized Mean Error (SME), Standardized Mean Absolute Error (SMAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE).

Results show that augmenting demand-only models with seat capacity and fare improves forecast performance for a substantial share of O–D markets. However, the relative benefit of endogenous versus exogenous treatment varies systematically with market characteristics and demand volatility. Endogenous specifications better capture interdependence and feedback dynamics, whereas exogenous specifications offer operational advantages when future capacity and pricing plans are available or can be credibly projected. Overall, the findings demonstrate the feasibility of scalable, network-wide forecasting frameworks that explicitly test variable-treatment assumptions and provide empirically grounded guidance for balancing predictive accuracy, operational feasibility, and interpretability in O–D demand forecasting.

Keywords: Origin-Destination (O-D), passenger demand, endogenous variables, exogenous variables, seat capacity, airfare; Seasonal Autoregressive Integrated Moving Average (SARIMA), Seasonal Autoregressive Integrated Moving Average with Exogenous Regressors (SARIMAX).

Share

COinS