Modelling of Return of S&P 500 Using the Non Linear Generalized Autoregressive Conditional Heteroscedasticity (NGARCH) Model

Trimono, Trimono
Damaliana, Aviolla Terza
Putri, Irma Amanda

Nusantara Science and Technology Proceedings
8th International Seminar of Research Month 2023

ARIMA Box-Jenkins is one of the most popular forecasting methods. ARIMA modeling requires a non-heteroskedastic care that shows constant residual variants. Awake, meaning residual residue from heteroscedastic ARIMA modeling (not constant). To overcome the problem of residual heteroskedasticity ARIMA used modeling volatility that is Generalized Autoregressive Conditional Heteroscedasticity (GARCH). GARCH is used to model the ARIMA residual variant which means symmetric. Some financial data has an asymmetric nature caused by the influence of good news and bad news. To accommodate these asymmetric properties, we use the Non-Linear Generalized Autoregressive Conditional Heteroscedasticity (NGARCH) volatility model which is the development of the GARCH model. This research applies NGARCH model using S & P 500 share price index data from January 1, 2019, until July 31, 2023 during active day (Monday-Friday). The purpose of this study, to find the best model NGARCH. The best model generated for S & P 500 stock price index data is ARIMA (1,0,1) NGARCH (1,1) because it has small AIC.