Abstract:
Traditional tourism demand analysis uses ordinary least squares or maximum likelihood methods to
estimate demand models like Box-Jenkins and State-Space, assuming that the parameters of the models
remain constant over the sample period. This assumption is too restrictive, as it does not allow for
behavioral changes of arrival of tourists over time. This study proposes a new methodology the
Generalized autoregressive conditional heterosedasticity model (GARCH) approach to tourism demand
modeling. This method relaxes the assumption of parameter constancy, and the behavioral change of
tourists over time is traced using a statistical estimator known as a Kalman filter. GARCH models permit
time-varying conditional covariances as well as variances, and the former quantity can be of substantial
practical use for both modeling and forecasting. The appropriateness of the GARCH approach to tourism
demand modeling is tested based on a data set of the tourist demand for Sri Lanka and estimated Mean
Percentage Errors(MAPE) are explained 9.7% 6% and 2% respectively