Modelação da procura turística em Portugal: A metodologia Box-Jenkins, os modelos Naïf e o modelo de alisamento exponencial

Abstract

This paper models tourism demand in Portugal and examines the forecast accuracy of several univariate models, such as the methodology of Box-Jenkins (1976), Naïf models and exponential smoothing models.
To measure tourism demand, the number of monthly overnight stays in the hotel establishments was chosen for the period from January 1979 to December 2009. The sample was divided into two parts: the first (from January 1979 to December 2005) was used to estimate the models and the second (from January 2006 to December 2009) to evaluate the forecast performance of the models considered. Except for Portugal, where the exponential smoothing model showed better results, for all other countries the Box-Jenkins model proved to be superior to the others as a forecasting model. Hence, these results suggest that the complexity of the models may result in improved forecast performance. In terms of the results of this study it was also observed that the Naïf models, based on yearly differences, presented superior results than when first (monthly) differences were considered, reinforcing in this way the importance of seasonality in tourism.