Forecasting Tourism Demand for Taiwan from the US, Canada and Europe

This paper examines five time-series forecasting models and fits them to the datasets of yearly, quarterly and monthly visitor arrivals in Taiwan from the United States, Canada and Europe. We then use the resulting estimated models to make ‘out-of-sample’ forecasts over the next 2 years (8 quarters and 24 months). We find that forecasts generated using the autoregressive integrated moving average and fractionally integrated autoregressive moving average models provide more accurate forecasts than those from other models when judged on the basis of mean absolute percentage errors and root mean square errors criteria. If there has been a structure break because of SARS outbreak in early march 2003, the result obtained by assuming a constant parameter structure during the entire period will be suspected. An analysis of the time series, including unit root tests with structural break is carried out as well.