Analysis And Short-term Forecast Of International Gold Spot Price Returns

With the continuous development and innovation of gold investment products,gold is becoming a mainstream investment tool due to the increase of economic and political uncertainties.The rapid development and globalization of economic and trade have made the gold market continue to expand in depth and breadth.The linkage between other markets has enhanced,and the drivers of gold prices have become increasingly complex.In the beginning of 2020,risk events occur frequently.The spread of 2019-n Co V led to a large-scale shutdown and then the market vitality decline.Increased market uncertainty and superposition of geopolitical risks have led to a plunge in oil prices and the fusing of the stock market.Concerns about the prospects for economic development have sent the market into a state of panic and downturn.Gold rise in an uncertain and fragile economic environment,which is a hedge asset during the crisis.Firstly,based on the existing literature and the analysis of the current market environment,this paper combs the driving factors of gold price,constructs a non Gaussian SVAR model to study the dynamic relationship among gold price returns,US dollar index volatility,crude oil price returns,VIX volatility and ten-year US Treasury bond yield volatility.Secondly,based on two perspectives,this paper makes a comparative analysis of the short-term forecast of gold price returns.One perspective is to build a Bayesian structure time series model based on the multi variable prediction of macroeconomic indicators.The driving factors of gold price and its first-order lag are included in the model,and study whether the effective prediction variables of gold price change in the whole sample period and financial crisis period;another perspective is the single variable prediction based on the historical data of gold price returns,construct exponential smoothing based on STL decomposition and neural network model,ARIMA and ETS models which are widely used and more robust in the literature are selected as the benchmark models.Make a short-term prediction of the gold price returns and compare the prediction accuracy,then select the prediction model which is most suitable for the gold price return rate during the sample period.The main research conclusions of the paper are as follows:(1)The gold price returns,crude oil price returns and the volatility of US Treasury bond yields have a positive impact on gold price returns,the US dollar index and the VIX volatility have a negative impact on the gold price returns.The impact of various variables on the gold price will change over time.Among them,the impact of the dollar index volatility has the largest impact,and the crude oil price yield impact has the longest duration.Overall,the duration of the impact is five trading days.(2)During the whole sample period,DEBT,DEBTL1,OIL,Oi L1,USDX and USDXL1 are the most important predictors of gold price returns.The current value of VIX Index's volatility can't effectively predict the gold price,but its first-order lag value has certain prediction ability.In the financial crisis period,except the dollar index and its first-order lag,other variables lose the prediction ability of the gold price yield.(3)This paper compares the short-term forecasting ability of five forecasting models to gold price returns,BSTS,STL-ETS,neural network,ARIMA,ETS.The results show that in the sample period selected in this paper,the Bayesian structure time series model has the highest forecasting accuracy to gold price returns.