HAR Model to Examine the Impact of Daily, Weekly, and Monthly Effect

Gold has always been crucial in economy in both ancient and modern societies and people are interested in investing in it all the times, whether in traditional physical gold or modern gold futures. In this paper, we want to use more advanced and scientific modern means to make prediction on the future volatility of gold futures to help investors make decisions and reduce risks. Based on heterogeneous autoregressive (HAR) theory, we establish three heterogeneous autoregressive realized volatility (HAR-RV) models to predict the future volatility of gold futures at three different horizons (daily, weekly, and monthly) utilizing 5-minute-frequency trading data in Chinese gold futures market from 01, 2008 to 02, 2021. The empirical result shows that our HAR-RV models is better at forecasting the future weekly volatility than future daily and monthly volatility, in terms of both statistical significance and level of goodness of fit. Also, the forecasted future volatility in daily, weekly and monthly HAR-RV models has a stronger relation with weekly, weekly and monthly realized volatility separately.