The Development and Testing of a Modified Diffusion Model for Predicting Tourism Demand

This study attempts to apply the Bass model to predict tourism demand. Price, income, and distribution channel of tourism are integrated into a general diffusion model of tourism industry. Empirical results for several modified Bass models are reported. Theoretical research results demonstrate that the univariate diffusion model remains robust with the inclusion of other variables for a variety of multivariate models. Alternative diffusion models were empirically assessed and compared by employing data for Taiwan's outbound travelers. Empirical research findings verify that price and income improve the diffusion model; distribution negatively affects prediction. Introduction Research has clearly established the central role of long- and short-term forecasting in tourism: short-term forecasts are required for scheduling and staffing by businessmen and leisure travelers (Brodie and Kluyver, 1987); and long-term forecasts are key to government investment in transportation and infrastructure (Wandner and Van Erder, 1980). After a broad review of tourism forecasting schemes (Lim, 2004) a change in emphasis from explaining forecasting approaches to discussing the application of diese techniques is desirable; however, the techniques used to forecast tourism demand differ dramatically. The majority research is concerned with econometric studies (Goh and Law, 2002; Lim and Mcaleer, 2002). Other current quantitative forecasting procedures have been adopted to predict tourism demand, including technical analysis techniques (Petropoulos et al., 2005), neural networks (Law, 2000), fuzzy theory and grey theory (Wang, 2004). The diffusion model has not, however, been explicitly explored in the tourism literature. Since diffusion theory was first introduced by Bass to marketing in 1969 (Bass, 1969), innovative diffusion theories have sparked research among consumer behavior, management, and marketing science areas other than tourism demand. Hereafter, the diffusion model is named the Bass model. Researchers who have used the Bass model to forecast tourism demand, are the exceptions. Morley (1998) used a logarithmic transformation of the raw data in the modeling of tourism demand. This study incorporated a variety of marketing variables such as channel, price, and income into the standard Bass model, employing data for Taiwan's outbound travelers and using established modified diffusion models. These variables integrated into the Bass model are important factors in forecasting durable goods demand. This study further evaluates whether or not these variables have the same influence in tourism demand as they do in other areas. The major purpose of this study is to identify whether these models fit the tourism industry and to compare the precision of these different models. Bass model Establishing the Bass theoretical model The purpose of mis study is to establish and evaluate the Bass model for predicting tourism demand. Some marketing variables appropriate to the tourism demand functions are incorporated into the model. Bass model with price variable Bass model with distribution variables Theoretically, population changes will affect the market potential of a destination. This study incorporated the variable of population into the Bass tourism model, using the dynamic concept proposed by Mahajan and Peterson (1978). Bass model with income variable Increases in travelers' income result in an improved quality of living, and greater motivation to travel. Consequently, the income function fitted with the exponential curve (Robinson and Lakhani, 1975), is IM - eψI(t) where ψ is the sensitivity of income coefficient, and I(t) is the per capita income of Taiwan at time t. Empirical research of the Bass model The data Time-series data for four variables in Taiwan were obtained. Two data sets were employed for the marketing mix and two sets were employed for the environmental variables. …