Emulating the southwest airline experience to forecast demand for High Speed Rail

Abstract

This research primarily deals with developing a new forecasting methodology that takes "induced traffic" into account, calibrates a relational model - attributing the growth in traffic in a particular market to factors such as decrease in relative fare and increase in relative frequency and finally applies a combination of the above two modules to obtain total forecasts for various market scenarios.

Ridership is a key determinant of financial feasibility for starting any new service irrespective of other system characteristics. Consequently, demand forecasting models play a vital role in any transportation planning process. The present research derives its motivation from previous efforts, wherein, noticeable deviations have been observed between actual and predicted ridership values. One of the primary reasons that is identified for these deviations is "induced traffic."

The overall goal of this project has been to develop a new approach for studying the effect of change in relative fare and relative frequency on variation in total ridership forecasts for High Speed Rail (HSR) in various market scenarios, by drawing from the experience of another common carrier having similar characteristics.

First, we study the "Southwest Airline Experience," which has been identified to be very close in character to HSR. For this carrier, the suggested methodology is used to derive the predicted traffic volumes and deviations from actual volumes are obtained. A data set of 33 city-pairs is built upon the quantification of the "extra traffic," which is a combination of "traffic diverted from rail plus auto plus bus" and "induced traffic." Data trends and analysis are presented next. A relational model is built between the dependent variable i.e., "ratio of extra to total traffic" and the independent variables i.e., "relative fare" and "relative frequency." Results of model formation and validation are incorporated. The R-square value for the model has indicated a fairly good explanatory relationship. Also, the validation conducted on five test cases showed very encouraging results.

Final analysis of effect of marginal change in total traffic due to marginal changes in relative fare and relative frequency for three city-pairs are also presented. It was seen that either decreasing the relative fare marginally or even increasing the relative frequency marginally tended to increase the total traffic substantially.