Abstract
"The tourism sector has been recognised as a global driving force for economic growth given its rapid development in recent decades. Thus the importance of accurate tourism demand forecasting has attracted increasing attention from both academics and practitioners, especially due to the perishable and seasonal nature of tourism demand. After reviewing the methodological development of the tourism demand forecasting literature, incorporation of spatial spillovers has been identified as one of the emerging trends in tourism demand forecasting. Following this new trend as well as some empirical evidence confirming the existence of spatial spillover effects in tourism demand across destinations, this thesis aims to develop a series of spatiotemporal econometric methods for tourism demand forecasting.
The format of this thesis is by manuscripts. The first paper is a comprehensive literature review, and the second and third papers are two empirical studies based on two newly developed spatiotemporal methods. The literature review article covers 72 tourism demand forecasting studies published in the past decade. From the literature review, spatial econometrics has been identified as a promising trend for future tourism demand forecasting studies. Following this new trend, the first empirical study proposes a local spatiotemporal autoregressive model, which fully reflects both spatial spillovers and spatial heterogeneity. This newly developed model is then applied to forecast international tourism demand in 37 European countries. The second empirical study is built upon the first empirical study, by further extending the spatial model into a general nesting spatiotemporal (GNST) model, which incorporates additional explanatory variables in both space and time, and represents the most general form of spatiotemporal model even beyond the tourism forecasting literature. Results in both studies have shown improved forecasting performance compared with the benchmark models. Both empirical studies represent the first attempts of proposing the methods in the literature, thus contributing to not just the field of tourism forecasting, but broadly the general field of forecasting. The findings of the empirical studies also make practical contributions to the tourism industry in a destination. To be more specific, the confirmed superiority of the two spatiotemporal models in the empirical studies indicate that where significant spatial spillover effects are identified, destinations should closely monitor the change in tourism demand as well as the economic situation of their neighbouring countries to develop timely strategies in response to the prediction of indirect impact on tourism demand generated from other destinations. Furthermore, with the built-in parameter optimisation process along with the GNST model, forecasters are able to identify the optimal specification from a range of potential parameter combinations based on the characteristics of the empirical case. As such, this research improves the flexibility and efficiency of tourism forecasting practice. "