Modelling and Forecasting Tourist Arrivals in Sri Lanka

Abstract

Tourism is one of the major industry shows a rapid growth in the Sri Lankan economy. According to the annual tourism statistics, the international tourist arrivals shows 4.4% growth in 2015 and 27.72% growth in foreign exchange earnings in the same year compared to 2014. Therefore, understanding and examining the upcoming trends of tourist’s arrivals is really important and it will be beneficial and important for stakeholders and interesting parties of the country. The purpose of this research study is to investigate and forecast the tourist’s arrival in Sri Lanka based on the available past data. The collected tourist’s arrivals data from 2000 January to 2014 December are used for this study. For the reason that the tourist’s arrivals data follow univariate time series, the time series techniques, ARIMA and GARCH models, are proposed to use in forecasting. Since the data consists with heteroscedasticity, transformation methods are needed to use in some time series modelling approaches. In GARCH model approach, original data is used for identifying a suitable model while Box-cox transform is used in SARIMA model approach to overcome the heteroscedasticity problem. Basically, model selection is done based on AIC values and MAPE, MAE and RMSE values used for measure the performance of selected models. Among the proposed time series models, none of the SARIMA models are fitted well for the data as they are not diagnostic. Finally, ARCH (1) model with optimal lag (2, 7, and 12) is identified as the best model to forecast the future values of tourist’s arrivals in Sri Lanka.