Predicting tourism recovery from COVID-19 : a time-varying perspective

Abstract

The uncertainties associated with the coronavirus disease 2019 (COVID-19) pandemic significantly reduced the accuracy of traditional econometric models in forecasting tourism demand, as the relationship between tourism demand and its determinants during the crisis changes over time. To address these inaccuracies, we apply three Factor mixed data sampling (MIDAS) models with different time-varying parameter (TVP) settings: Factor TVP-MIDAS, Factor MIDAS with stochastic volatility (Factor MIDAS-SV), and Factor TVP-MIDAS-SV. We examine the dynamic relationship between tourism demand and its influencing factors, capture the uncertainty and volatility in the data, and provide short-term forecasting and nowcasting. We expose the Factor MIDAS models with TVP specifications to different combinations of determinants to examine their performance. The empirical results show that the Factor MIDAS models with TVP settings performed better than the Factor MIDAS model in the short-term forecasting and nowcasting of tourism demand during COVID-19. The results also suggest that high-frequency data complement these Factor MIDAS models with TVP settings in improving the forecasting and nowcasting accuracy during crises.