Bayesian models for tourism demand forecasting

Abstract

This study extends the existing forecasting accuracy debate in the tourism literature by examining the forecasting performance of various vector autoregressive (VAR) models. In particular, this study seeks to ascertain whether the introduction of the Bayesian restrictions (priors) to the unrestricted VAR process would lead to an improvement in forecasting performance in terms of achieving a higher degree of accuracy. The empirical results based on a data set on the demand for Hong Kong tourism show that the Bayesian VAR (BVAR) models invariably outperform their unrestricted VAR counterparts. It is noteworthy that the univariate BVAR was found to be the best performing model among all the competing models examined.