FRACTIONALLY-INTEGRATED AND MARKOV-SWITCHING MODELS TO FORECAST THE DEMAND OF A FAST-GROWING EMERGING AIR TRANSPORTATION MARKET

Abstract

This paper presents a methodology using a portfolio of time-series models, from conventional ARMA to a regime-changing framework. The objective is to develop an air transportation demand modeling to inspect potential structural breaks in the Brazilian market, due to solve underlying issues of new demand creation. Out-of-sample forecasting is used to generate comparative metrics aiming at both selection and validation of what we call a “champion model”. Results indicate better performance of more complex models such as the fractionally integrated and the Markov-switching models. Ex ante knowledge of the interaction of structural breaks and unit root can prove useful in the modeling analysis. The demand forecast of the champion model is in line with the recent accelerated growth of the Brazilian air transportation market, roughly 7% per year.