Forecasting the Demand for Tourism in China by Applying the Skew-<i>t</i> Distribution Model

This study investigates the demand for tourism in China using ARMA models with a normal distribution and with a skew-<i>t</i> distribution (referred to as ARMA-N and ARMA-skew-<i>t</i> distribution models). The empirical sample was selected with a view to discussing and comparing the ability of these models to accurately forecast the out-of-sample demand for tourism in China. The results indicate that the ARMA-skew-<i>t</i> distribution model is superior to the ARMA-N model in terms of forecasting the demand for tourism in China from the five countries considered in the study, based on mean squared errors (MSE) and mean absolute errors (MAE) tests. The Diebold-Mariano (DM) test further confirms that in all cases the ARMA-skew-<i>t</i> distribution model is more accurate in forecasting tourism demand in China than the ARMA-N model. These findings demonstrate the significant impact of both skewness and tail-thickness on the conditional distribution of returns.