Forecasting Seasonal Demand for Tourism in Iran: Application of Time Series Techniques

Introduction
Tourism industry plays a major role in creating job opportunities and income generation in all countries and it has grown remarkably in recent decades. Iran has a unique situation in tourism industry due to its amazing ancient monuments and natural attractions. Therefore, developing the tourism industry can be a suitable way to improve Iranian economy and can reduce its dependence on oil income. The purpose of this paper is modeling and forecasting seasonal flows of tourist arrivals into Iran. Moreover the forecasting accuracy of methods is compared. There are different methods that can be used to forecast the economic variables. Today forecasting is regarded as an important instrument for economic policymakers.
 
Materials and Methods
In order to deal with seasonality, Autoregressive Integrated Moving Average Approach (ARIMA) processes have been generalized. When modeling time series with systematic seasonal movements, Box and Jenkins recommend the use of Seasonal Autoregressive (SAR) and Seasonal Moving Average (SMA) terms. Therefore, we utilized the Seasonal Autoregressive Integrated Moving Average Approach (SARIMA) and seasonal integration model based on seasonal unit root test. We used Hylleberg et al. HEGY test is used for unit root testing. HEGY developed separate regression based T and F tests for unit roots at various frequencies in the quarterly data.  
The seasonal data about the number of tourist arrivals to Iran was obtained from the Iranian Cultural Heritage and Tourism Organization. Time scope covers 44 seasons, from 2001 to 2010.
 
Discussion and Results
Quarterly series of tourist arrivals shows the periodic behavior. HEGY test results indicate the presence of non-stationary tourist arrivals series. Therefore SARIMA and seasonal integration models are fitted into the data. In order to achieve stationary series, these time series need to be seasonally differentiated. In the next step, SARIMA and seasonal integration models were estimated. ARIMA (1,1,0) (1,1,1)₄ model identified as the best model among the SARIMA candidates. Finally, two indicators including RMSE (root mean squared error), MAPE (mean absolute percentage error) were employed in order to measure the performance of models. 
 
Conclusion
This study has discussed about two kinds of seasonal models including seasonal autoregressive integrated moving average approach (SARIMA) and seasonal integration model. Results of HEGY’s seasonal unit root test demonstrated that seasonality unit root exists in the tourist arrivals to Iran. Furthermore, the comparison of forecasting accuracy revealed that seasonal integration model has high accuracy more than seasonal ARIMA model. Thus, the seasonal integration model was selected as best model to forecast of tourism arrivals to Iran. This result is important to decision makers to evaluate tourism arrivals.