Modelling and Forecasting International Tourism Demand – Evaluation of Forecasting Performance

Abstract

The paper examines the forecasting accuracy of different forecasting techniques in modelling and forecasting international tourism demand in Croatia. As tourist arrivals is the most commonly used measure of international tourism demand, the realized number of German tourists arrivals in the period from first quarter of 2003 to the last quarter of 2013 is taken as a measure of tourism demand in Croatia. In this paper following forecasting techniques are compared: the seasonal naïve model, the Holt-Winters triple exponential smoothing, the seasonal autoregressive integrated moving average model (SARIMA) and the multiple regression model. After approaching the forecasting procedure, all models are compared considering the in sample and the out of sample mean absolute percentage error (MAPE). All compared models show good forecasting performances. Although the diagnostics for the selected models reveals that the four models do not significantly differ, it can be concluded that multiple regression model perform a highly accurate forecasting of German tourists arrivals in Croatia.