An application of PAR models for tourism forecasting

This paper extends the existing literature on tourism forecasting by developing an application study based on periodic models. The need for this type of approach derives from the common procedure in tourism literature of classifying series into three seasons: peak, shoulder and off-peak. This classification is useful in identifying more parsimonious periodic models when compared with unrestricted periodic monthly models. The authors apply their proposed models on several tourism series collected and analysed from Portugal's southernmost province, Algarve. The economy of this region relies heavily on the tourism industry, catering largely for the European market. Besides statistically validating these models, the authors compare their forecasting ability with other models in the current literature. The results show that the models presented achieve a superior forecasting performance.