| As currency abuse continues to erode the credit of the US dollar,gold has seen an important opportunity to develop as the main reserve asset for de-dollarisation.Despite being recognised as one of the preferred financial assets for international capital hedging,gold itself is not absolutely safe.Under the influence of various complex factors such as trade protection,fringe politics,finance and nature,financial losses arising from the volatility of gold market risks can be far higher than in other markets.The price of gold is highly volatile,often with an amplitude of more than ten dollars per ounce during the day,and even more so in the event of Black Swan events and the impact of interest rate hikes on the US dollar.On 11 August 2020,the single day amplitude of the gold price was a record high of $129.2 per ounce.With such large price fluctuations,how to scientifically measure the risk of the gold market,build a solid risk prevention and control system,and maintain the safe and smooth operation of the market,with a view to reducing the possibility of financial systemic risks arising from the gold market,is an important issue that needs to be addressed urgently.In the risk measurement of the gold market,it was found that too many factors interfere with the gold price,making the distribution of gold price returns differ significantly from the traditional normal distribution,with features such as spikes and thick-tailed left skew.The risk measurement model based on the assumption of the traditional normal distribution then tends to ignore such features when measuring gold market risk,affecting the accuracy of the risk measurement.We use the FZ0 loss function to circumvent the sample distribution assumptions and combine it with the Generalized Autoregressive Score(GAS)model to construct a time-varying semi-parametric model that highlights the characteristics of the left-hand side of the spike and the thick tail of the sample to conduct an empirical study in order to better reflect the market risk of the gold market.In the empirical study,the daily log returns on the closing price of gold futures on the New York Mercantile Exchange(COMEX)from May 10,1994 to July 30,2021 were selected for the empirical study,with a total of 6805 data.From the basic statistical information of the sample,it is found that the sample has characteristics such as spikes,thick tails and left bias.This paper then uses ten models such as rolling window model,GARCH model and GAS model to forecast VaR and ES under different confidence intervals in the gold market and compares the average losses generated by different models.Goodness-of-fit and DM tests are introduced to compare the ten models used to measure risk in the gold market.Finally,by empirically validating several other metal markets with spiky thick-tailed distributions,it is again verified that the GAS semi-parametric model proposed in this paper has good predictive power for VaR and ES and can be used for risk measurement in gold markets. |
PDF Full Text Request