Improving the forecasting of inbound tourism demand based on the mixed-frequency data sampling approach: evidence from Australia

Abstract

ABSTRACT This study explores how the mixed-frequency data sampling (MIDAS) approach enhances the forecasting of Australia’s inbound tourism demand by employing an autoregressive distributed lag (ARDL)-MIDAS model. The main findings are as follows: First, after capturing the effects of control variables, both daily exchange rate returns and daily exchange rate volatility affect Australia’s inbound tourism demand. Second, the monthly growth rate of inbound tourist arrivals follows a mean-reverting process and incorporating its historical fluctuation information from the past 3 months significantly increases the explanatory power of the ARDL-MIDAS model. Third, the results of the out-of-sample predictive performance indicate that the two MIDAS-based models significantly outperform the benchmark model and the other two candidate models due to the incorporation of intra-month exchange rate information. These findings provide insights into the forecasting of inbound tourism demand and lay the foundation for further tourism business planning, resource allocation, and policymaking.