As a safe-haven investment and a place to store value in times of economic uncertainty and market volatility, gold plays an essential function. Furthermore, the volatility in the gold futures market can be an indicator of the mood of the market and economic circumstances. For instance, higher levels of volatility might indicate increased global tensions, market uncertainty, or rising inflation pressures (Asai in 2020; Dutta et al., 2021, Yao and others. 2021). Monitoring the volatility of gold futures closely can help to modify the investment choices of institutional investors as well as portfolio managers. Knowing the patterns of volatility and trends in the gold futures market can help in determining the best investment strategies and the best way to allocate capital (Balcilar et al. in 2017. Demirer and co. 2019, 2019. Wei as well. 2020).

Increases in the price of gold could signal a change in the market’s dynamics or beginning of an emerging trend. Furthermore, gold typically shows correlations to other markets including equities, currencies and bonds. The price of gold’s rise could have an impact on these closely linked markets. The analysis and monitoring of gold price fluctuations can assist investors and traders to anticipate possible movements in market correlated to it and adjust their positions in line with. This information is useful in terms of diversification and managing the risk of portfolios.

This paper is similar to Luo et al. (2016) that examines the content of information in implied volatility as well as jumps in forecasting the realized volatility of Shanghai gold futures. Our paper differs from their research by taking into account the use of a different benchmark model. Luo et al. (2016) apply the autoregressive model while we consider the heterogeneous autoregressive-realized volatility (HAR-RV) model.

The paper’s contributions include the following. We first determine if the jump component that is constructed from high-frequency gold price data is useful for forecasting the fluctuations of the Chinese markets for gold by using the HAR-RV model. Furthermore, we discover that the jump component data can still be useful in predicting the volatility of gold futures under an alternative forecasting window as well as another test for evaluation.

Section Snippets

Realized volatility

In accordance with Andersen as well as Bollerslev (1998) In the work of Bollerslev and Andersen (1998), the daily realized volatility may be:RV==1,2, where Rt,j refers to the jth day’s intraday return of the day t, and M is the number of observations.

Following Barndorff-Nielsen and Shephard (2004), when D – 0, RVt can be:RV-02()+0<<=2(),where 02() is the integrated variance that can be computed by the bi-power variance (BPV), which is:BPV=1-2=21/|,||,-1|,where 1=2/0.7979. 2 = 0() can be described as the non-continuous portion of the quadratic

Data

The high-frequency prices for gold futures on Shanghai Futures Exchange (SHFE). Shanghai Futures Exchange (SHFE) are derived via Wind database. Wind database. It was officially launched on April 19, 2016. Table 1 provides the statistical information.

Analytical evidence

In accordance with Patton (2011) In accordance with Patton (2011), we use using the QLIKE as well as the MSE loss function to carry out an out-of-sample analysis. These two loss functions are defined as follows:QLIKE=-1=1((RV^)+RVRV,MSE=-1=+1(RVm-RV^)2,where RV^ indicates prediction from the forecasting model, while RVm is the actual volatility. L refers to the length outside of sample.

We rely on the combination of models confidence sets (MCS) from Hansen and co. (2011) as well as the loss functions to select the most appropriate model set. The confidence level of a = 0.25. Moreover,

Another test for evaluation

Table 3 summarizes the results of the test that was conducted outside of the sample 2 test. 2. test. When you find that the OOS2 value of a model is greater than zero, it indicates that the model has an advantage over that of the model used as a benchmark. From Table 3, we see an OOS2 value of HARRV-J is 1.2066 and it shows information about jumps. can increase the accuracy of forecasting for the HAR-RV model.