An assessment of combining tourism demand forecasts over different time horizons

Abstract

This study investigates the performance of combination forecasts in comparison to individual forecasts. The empirical study focuses on the U.K. outbound leisure tourism demand for the United States. The combination forecasts are based on the competing forecasts generated from seven individual forecasting techniques. The three combination methods examined in this study are the simple average combination method, the variance–covariance combination method, and the discounted mean square forecast error method. The empirical results suggest that combination forecasts overall play an important role in the improvement of forecasting accuracy in that they are superior to the best of the individual forecasts over different forecasting horizons. The variance–covariance combination method turns out to be the best among the three combination methods. Another finding is that the encompassing test does not significantly contribute to the improved accuracy of combination forecasts. This study provides robust evidence for the efficiency of combination forecasts.