Forecasting of gold price return volatility using a nonparametric GARCH model and compare with parametric GARCH models

In recent years, investment in gold has been remarkable for investors because of a recession in stock exchange. This increase in demand of gold caused increase in gold price. Because of increase in gold price, dealing of gold expanded and so volatility of gold price return increased intensly. So we have to use a model to predict volatility beside return to make decision for investment.
Find a model that it can do a better forecast of price return volatility is a debatable topic in the finance literature. Around this topic some models have been presented and these models have some advantages and disadvantages. These models have been applied for predict of volatility of crude oil and exchange rate more than other fields. Between all models, GARCH models have been more applicable than others. So we use this group of models too, but in a different way. This way is a nonparametric approach to GARCH model that presented by Buhlman and McNeil for first time in 2002. In this research we use this approach to forecast volatility of gold price return and compare it with other GARCH models by two loss function (QLIKE-MSE). The result of this research shows that nonparametric GARCH has a better performance than the other GARCH models based on QLIKE loss function with a statistical significance, but based on MSE loss function we can’t judge.