GLOSTEN JAGANNATHAN RUNKLE-GENERALIZED AUTOREGRESSIVE CONDITIONAL HETEROSCEDASTICS (GJR-GARCH) METHODE FOR VALUE AT RISK (VaR) FORECASTING
A stock returns data are one of type time series data who has a high volatility and different variance in every point of time. Such data are volatile, seting up a pattern of asymmetrical, having a nonstationary model, and that does not have a constant residual variance (heteroscedasticity). A time series ARCH and GARCH model can explain the heterocedasticity of data, but they are not always able to fully capture the asymmetric property of high frequency. Glosten Jaganathan Runkle-Generalized Autoregresive Heteroskedascticity (GJR-GARCH) model overcome GARCH weaknesses in capturing asymmetry good news and bad news taking into the leverage effect. Furthermore GJR-GARCH models were used to estimate the value of VaR as the maximum loss that will be obtained during a certain period at a certain confidence level. The aim of this study was to determine the best forecasting model of Jakarta Composite Index (JSI). The model had used in this study are ARCH, GARCH, and GJR-GARCH.