ABSTRACT
This chapter proposes a demand forecasting method based on the stochastic frontier analysis (SFA) method and a model average technique and describes an empirical application of the medium-term air travel demand forecasting. It estimates the underlying demand with the historical passenger traffic and several explanatory variables of air travel demand. SFA is a parametric method that was developed to estimate efficient frontier and efficiency scores. Additionally, considering uncertainty in the selection of SFA models, cross-validation type model averaging is applied to improve the forecasting accuracy. Model averaging, that is, setting weights to candidate models in some way, can prevent one from putting all of his/her inferential eggs in one unevenly woven basket. Model averaging often reduces the risk in regression estimation, as 'betting' on multiple models provides a type of insurance against a poor singly selected model.
