In this paper we investigate the long memory of Tehran Securities Price Index and fit ARFIMA model using 970 daily data since 1382/1/6 until 1386/4/17. Furthermore, we compare the forecasting performance of ARFIMA and ARIMA models. The results show that the series is a long memory one and therefore it can become stationary by fractional differencing. We obtaine the fractional differencing parameter . Having done the fractional differencing and determination of the number of lags of autoregressive and moving average components, the model is specified as . We estimate the parameters of the model using 900 in-sample data and use this estimates for forecasting 70 out-sample data. Comparing forecasting performance of two models illustrate that forecasting performance of ARFIMA model is better than ARIMA model.
JEL Classification: A12