Modeling Jakarta composite index with long memory and asymmetric volatility approach

Abstract

Jakarta Composite Index (JCI) is a value that shows the performance of all listed shares on the stock exchange in Indonesia. JCI also describes the condition of investment, whether it is strong or weak. Therefore analysis and forecast related to JCI are the important things for investors. Most of the economic data have a strong correlation even though the observation distance is far, or called long memory process. Because of its high fluctuation in the stock market, the JCI also has high volatility. As an impact of high volatility, positive and negative signs gave different effects on volatility movements that called asymmetric effects. The data used in this study is JCI in return for 2007-2018, which is expected to have a long memory effect and high volatility. ARIMA-GARCH, ARIMA-EGARCH, ARFIMA-GARCH, and ARFIMA-EGARCH are applied for modeling the data. ARIMA-GARCH and ARFIMA-GARCH are models for data without asymmetric effect, while ARIMA-EGARCH and ARFIMA-EGARCH are models for data with asymmetric effect. The tests conducted to see the asymmetric effect are Sign Bias Test, Positive Sign Bias Test, Negative Sign Bias Test, and Joint Effect Test. The conclusion in this study is that JCI is a long memory and has an asymmetric effect. The join effect test in asymmetric effect test is significant in all models formed, which means positive and negative sign give different effects on volatility movements simultaneously. Based on the RMSE value, the best model for forecasting JCI is the ARFIMA-EGARCH model.