Tourism Forecasting Methods: An Overview

Tourism forecasts may be generated by either quantitative or qualitative approaches. Quantitative approaches develop and employ mathematical models, theories and hypotheses pertaining to natural phenomena. Unlike quantitative approaches, qualitative approaches involve in-depth understanding of human behaviour and the reasons behind various aspects of behaviour. Simply put, qualitative approaches investigate the why and how of decision making, as compared to what, where, and when of quantitative approaches.

Quantitative Approaches

Studies on tourism demand analysis using quantitative approaches fall into two major groups: casual (econometric) models and non-casual (mostly time series) techniques.

• Econometric models: This approach uses regression analysis to estimate the quantitative relationship between tourism demand and its determinants. For detailed discussions on modern econometric approaches, please refer to a book, Tourism Demand Modelling and Forecasting (written by Haiyan Song and Stephen Witt).

• Time series techniques: This approach involves extrapolate historic trends in tourism demand into the future without considering the underlining causes of the trends. The most frequently used methods in this approach include Exponential Smoothing and Box-Jenkins procedure.

Qualitative Approaches

Studies on qualitative forecasting in the tourism field are very limited. Most of these studies have focused on Delphi method and scenario analysis.

Forecasting Method Used In This Research

This research focuses on quantitative forecasting methods, and in particular, econometric ones. Specifically, a vector autoregressive (VAR) model has been employed. The VAR approach models the econometric relationships using a system of equations in which all the variables are treated as endogenous.

The reasons for using VAR model in this research are as follows:

1. VAR model can generate relatively accurate medium and long term forecasts of tourism demand;

2. VAR model does not require the generation of forecasts for the explanatory variables before the forecasts of the dependent variable can be obtained;

3. Impulse response analysis can be carried out, which can provide useful information for policymaking purposes.