The tourism industry has become one of the fastest growing industries in the world, with international tourism flows in year 2006 more than doubled since 1980. In terms of direct economic benefits, United Nations World Tourism Organization (UNWTO, 2007) estimated that the industry has generated US $735 billion through tourism in the year of 2006. Through multiplier effects, World Travel and Tourism Council (WTTC, 2007) estimated that tourism will generate economic activities worth of approximately US $5,390 billion in year 2007 (10.4% of world GDP). Owing to the important economic contribution by the tourism industry, researchers, policy makers, planners, and industrial practitioners have been trying to analyze and forecast tourism demand. The perishable nature of tourism products and services, the information-intensive nature of the tourism industry, and the long lead-time investment planning of equipment and infrastructures all render accurate forecasting of tourism demand necessary (Law, Mok, & Goh, 2007). Past studies have predominantly applied the well-developed econometric techniques to measure and predict the future market performance in terms of the number of tourist arrivals in a specific destination. In this chapter, we aim to present an overview of studies that have adopted artificial intelligence (AI) data-mining techniques in studying tourism demand forecasting. Our objective is to review and trace the evolution of such techniques employed in tourism demand studies since 1999, and based on our observations from the review, a discussion on the future direction of tourism research techniques and methods is then provided. Although the adoption of data mining techniques in tourism demand forecasting is still at its infancy stage, from the review, we identify certain research gaps, draw certain key observations, and discuss possible future research directions.
Econometric modeling and forecasting techniques are very much highly exploited to the understanding of international tourism demand. The econometric techniques have evolved and improved in accuracy and sophistication in recent years with the development of stationarity consideration, cointegration, and GARCH and time varying parameter (TVP) models. However, the development of such techniques might have reached a plateau for the time being. It is often expressed that econometric models are expanded at the expense of model comprehensiveness. That is, demand relationship is represented in a complicated system of equations in such models, which policy makers and practitioners find it like a black box that is hard to comprehend.
Besides, the conventional econometric models are founded on strict statistical assumptions and stringent economic theory (e.g., utility theory and consumption behavioral theory). These theories suggest that economic factors, such as income, price, substitute price, and advertising, are the primary influences of demand. However, tourism consumption, particularly if it involves a long-haul trip, is restricted by not only income constraint, but also on time constraint (Cesario & Knetsch, 1970; Morley, 1994), and the intrinsic properties of a particular tourism product or service that have not been modeled in the traditional demand framework (Eymann & Ronning, 1992; Morley, 1992; Papatheodorou, 2001; Rugg, 1973; Seddighi & Theocarous, 2002). The noneconomic factors, such as psychological, anthropological, and sociological factors, had not been analyzed in the traditional econometric travel demand studies. It is true that there exists a large amount of tourism literature that studies how these noneconomic factors could affect travel motivation, and how travel motivation affects destination choices (e.g., Edwards, 1979; Gray, 1970; Mayo & Jarvis, 1981; Um & Crompton, 1990), published articles have failed to show how noneconomic factors could affect demand for tourism, however.
Key Terms in this Chapter
Perishable Tourism Product: Tourism services are perishable, and cannot be stored for sale at a later date. This affects the distribution of tourism services, as they must be marketed in a way that minimizes lost capacities.
Cointegration: An econometric property of time series variables. If two or more series are themselves nonstationary, but a linear combination of them is stationary, then the series are said to be cointegrated.
Econometrics: It is concerned with the tasks of developing and applying quantitative or statistical methods to the study and elucidation of economic principles. It combines economic theory with statistics to analyze and test economic relationships.
Multiplier Effect: In economics, it refers to the idea that an initial spending rise can lead to even greater increase in national income
Statistical Assumptions: Statistical assumptions are general assumptions about statistical populations. In order to generate interesting conclusions about real statistical populations, it is usually required to make some appropriate background assumptions in order to generate valid and accurate conclusions.
Nonstationary: A condition where value of the time series changes over time, usually because there is some trend in the series so that the mean value is either rising or falling over time.