Forecasting Inbound Canadian Tourism: An Evaluation of Error Corrections Model Forecasts

Abstract

This paper computes and evaluates a variety of quantitative forecasts for inbound Canadian tourists, including the Error Corrections Model (ECM) and the traditional regression model forecasts. A number of forecasting methods are employed: naive to sophisticated, univariate to multivariate, time series and econometric. Forecasts for the number of inbound Canadian tourists are derived using data from four major markets: the USA, the UK, Germany and Japan. The evaluation of the forecasts is based on the Generalized Forecast Error Second Moment (GFESM) criterion developed by Clements and Hendry (1993) and the Adjusted Mean Absolute Percentage Error (AMAPE) criterion. The ECM forecasts performed best, while the traditional regression model forecasts performed poorly. In this study, using Canadian data, the development of an ECM (which entails careful analysis of the integration and co-integration properties of the variables) provides an improvement in forecast accuracy. Previous tourism studies have found less promising results concerning the performance of the ECM forecasts.