Modeling International Tourism Demand: Model Specification and Structure

Abstract

Multiple regression is used widely as a tool for estimating tourism demand functions. Problems of heteroscedasticity, multicollinearity, and autocorrelation are well recognized by modelers, but questions ofmodel specification are less widely understood. Misspecifi cation of the model, such as failure to include an important explanatory variable or a wrongfunctionalform, can have significant impacts on the model estimated. The evidence to date is that more complex or general functional forms give better models than more simple functions such as the commonly used log-linear model. It is reasonable to expect that tourism demand is a dynamicfactor, so the lag structure oftourism demand should be considered in any model. A quadratic functional form for this structure is derived.