Forecasting New Zealand Tourism Demand with Disaggregated Data

This paper compares the forecasting performance of the ARIMA model and the Winters Exponential Smoothing method against each other and the naive No Change process. The models are fitted to quarterly international tourist flow data to New Zealand, from June 1978 to September 1992. Forecasting performance is compared between 11 different countries and world regions with the travel flow divided by type of tourism into categories including Holiday travel, VFR travel and Business travel. It is concluded that the Winters and ARIMA methods outperform the No Change process. In all cases, the relative performance between ARIMA and Winters is affected by whether tourism is disaggregated by types of tourist travel, or analysed only as total flow.