Border Region Bridge and Air Transport Predictability

Thomas M. Fullerton, Jr.¹, Somnath Mukhopadhyay^2
(1. Department of Economics & Finance, University of Texas at El Paso, El Paso, TX 79968-0543, USA;
2. Department of Information & Decision Sciences, University of Texas at El Paso, El Paso, TX 79968-6282, USA)

Abstract: Border region transportation forecast analysis is fraught with difficulty. In the case of El Paso, Texas and Ciudad Juarez, Chihuahua, Mexico, dual national business cycles and currency market fluctuations further complicate modeling efforts. Incomplete data samples and asymmetric data reporting conventions further confound forecasting exercises. Under these conditions, a natural alternative to structural econometric models to consider is neural network analysis. Neural network forecasts of air transportation and international bridge activity are developed using a multi-layered perceptron approach. Those out-of sample simulations are then compared to previously published forecasts produced with a system of simultaneous econometric equations. Empirical results indicate that the econometric approach is generally more accurate. In several cases, the two sets of forecasts are found to contain complementary information.

Key words: regional transport demand; neural networks; econometric forecasting

JEL codes: R15, R41