Precious Metal Volatility Prediction Based On Improved GARCH-MIDAS Mode

In the complex and changeable macroeconomic environment,precious metals with the function of preserving value,increasing value and avoiding risk play an increasingly important role in investment risk management.If investors can effectively track the fluctuation of precious metal market in real time and optimize the investment portfolio strategy in time,market risks can be avoided.In order to improve the prediction accuracy of precious metal volatility model,it is worthwhile to explore the factors influencing precious metal price fluctuation.At present,volatility model based on mixed-frequency data is widely concerned.Among them,the generalized autoregressive conditional heteroscedasticity mixed-frequency data sampling model(GARCH-MIDAS,or GM model)is widely used.Some scholars put forward a GMAE model with asymmetric effect and threshold effect for the GM model without considering the asymmetric effect and threshold effect of the asset return series;Some scholars proposed the GM-SK model with time-varying skewness and time-varying kurtosis in view of the GM model ignoring the deficiency of skewness and peak fat-tailed of the asset return series.However,GM-AE model does not consider the influence of skewness and peak fat-tailed of asset return series on the model prediction performance,and the GM-SK model does not consider the asymmetric effect and threshold effect of asset return series.Due to the high complexity of precious metal market,in order to fully extract the effective information of precious metals market for prediction,we simultaneously consider the asymmetry effect,threshold effect,skewness and peak fat-tailed characteristics in precious metals series.Therefore,on the basis of GM-AE model,we constructed GM-AE-SK model by introducing time-varying skewness and time-varying kurtosis,and the realized volatility(RV),China Comprehensive Leading Indicator(CCLI)index and China Economic Policy Uncertainty(CEPU)index were used as low-frequency independent variables of GM-AE-SK model to analyze the impact of macroeconomic indicators on the prediction performance of precious metal volatility model.In this thesis,the gold,platinum,palladium and silver series and CCLI index and CEPU index in recent 16 years are selected as empirical data sets to empirically analyze the GMAE-SK model under different low-frequency independent variables from the perspective of in-sample estimation and out-of-sample prediction.The empirical analysis results show that:(1)GM-AE-SK model has better fitting performance and prediction performance than GM-AE model and GM-SK model in predicting precious metal volatility.(2)The prediction performance of GM-AE-SK model with CEPU index as low-frequency independent variable is better than that of GM-AE-SK model with RV index and CCLI index as low-frequency independent variable,indicating that CEPU index can improve the prediction performance of precious metal volatility model.(3)By examining the realized utility of various volatility models,it is found that the GM-AE-SK-CEPU model,which combines time-varying skewness and kurtosis with CEPU index,provides significantly improved realized utility,which is a better volatility prediction model,and can better predict the future trend of precious metals volatility,which can provides reference suggestions for investors and policy makers and regulators to understand the volatility characteristics of precious metals market.