Evaluating Time-Series Models to Forecast the Demand for Tourism in Singapore

Abstract

The authors look at eight models to forecast inbound tourist arrivals to Singapore, six of which were analyzed by Chan and by Chu. The authors explore model performance from a different perspective than either of these authors and arrive at different conclusions. Major suggestions are as follows: (1) a complete comparison among competing models during the estimation phase and a battery of performance statistics when evaluating these models sheds light on several top-performing models; (2) when evaluating the forecasting performance of competing models, different performance statistics may lead to different model selections; (3) among competing models, a model that performs best during the within-sample period does not necessarily perform best in the postsample period; (4) changing the length of the forecast horizon can have an effect on the choice of the best model; and (5) a combined model may be the one that provides the best forecasting performance.