Trading Gold Future with ARIMA-GARCH model

Nop Sopipan

Abstract

In this paper, we forecast volatility of gold prices using ARIMA-GARCH models. All models are estimated under three distributional assumptions which are Normal, Student-t and GED. The gold price log returns are stationary. We found that the ARIMA(2,0,2) gave the best performance model for forecasting the return of gold. Serial correlation in the squared returns suggests conditional heteroskedasticity. This empirical part adopts GARCH models to estimate the volatility of the gold price. To account for fat-tailed features of financial returns, we consider three different distributions for the innovations. The trading details we have used describe forecasts of a closed price of gold price and trading in the gold future contract (GF10J16). We found that the cumulative of return with ARIMA(2,0,2)-GARCH-N model and the ARIMA(2,0,2)- GARCH-GED model give cumulative of return more than the ARIMA(2,0,2)-GARCH-t models.