Volatility Forecasting And Application In Chinese Commodity Markets Based On The Global Stock Market Volatility Information

Commodities,as strategic raw materials and consumer goods for global industrial and agricultural production,are important guarantee for maintaining the stable operation of the economy,and unanticipated large fluctuations in commodity prices will inevitably impact the smooth and healthy development of the national economies.As global economic integration and the financialization of commodities continue,traditional commodities are gradually exhibiting financial attributes and the commodity market is becoming an important source of diversified returns for investors.Meanwhile,commodity price volatilities are increasingly showing a tendency to be closely linked to the financial markets on an international scale.Against this backdrop,price volatility in commodity markets is increasingly attracting the attention of leading academics,institutional investors and policy makers.Besides,with China’s economic enters the new normal stage,China faces a more severe external environment and more challenging development tasks.Therefore,modeling and forecasting the volatility of the domestic commodity market in combination with information on international stock market volatility is of great implication for gaining insight into and preventing risks in China’s commodity markets,improving commodity pricing power and resource allocation capabilities,and achieving high-quality development of China’s economy.However,the volatility of commodities has increased in recent years due to the financial crisis,speculative factors and supply/demand imbalances,making the modeling and forecasting of commodity volatility a challenging task.Against this background,this thesis focuses on the following questions: Are international stock markets indeed drivers of domestic commodity volatility? How does the intensity and direction of volatility information transmission between domestic commodity markets and international stock markets change over time and in the frequency domain? If it is true that volatility information can be transmitted from the international stock markets to the domestic commodity markets,does this imply that the volatility information of the international stock markets can predict the volatility of the domestic commodity markets? Can Markov regime switching techniques help to improve the prediction accuracy of high-frequency volatility in domestic commodity markets by making full use of the volatility information of individual international stock markets? What’s the difference between the ability of Markov regime switching technique with fixed and time-varying state transfer probabilities in improving the prediction accuracy of commodity volatility? Whether the shrinkage models based on Markov regime switching techniques and MIDAS-RV model can make better use of the volatility information of multiple international stock markets and produce higher prediction accuracy for the high-frequency volatility of domestic commodity markets than the traditional combination forecast methods and dimension reduction techniques?Firstly,this thesis discusses the dynamic volatility information transmission characteristics between the domestic commodity markets and the international stock markets,using the Kalman filter-based dynamic spillover index framework and its frequency domain extensions,and specifies the main volatility information transmission directions between the two types of markets.The results of the study find that there is mainly a net volatility spillover direction between the two types of markets from the international stock markets to the domestic commodity markets,with the international stock markets being the important dominant factors in the price volatility of domestic commodities.However,since commodities are usually the important raw materials and consumption for industrial and agricultural production,sometimes commodities also dominate the price volatility of the international stock markets.Besides,the occurrence of major economic and political events tends to strengthen the total volatility spillover intensity from international stock markets to domestic commodity markets,especially the intensity of the medium-and long-term total volatility spillover.Finally,there are phase differences in the frequency domain of the volatilitydominated relationships between international stock markets and some domestic commodity varieties.Secondly,this thesis uses the MIDAS-RV model as the benchmark and uses the MIDASRV-X model to discuss the predictive ability of international stock market volatility information for the high-frequency volatility of domestic commodity markets.The out-ofsample forecasting assessment results indicate that international stock market volatility information has statistically significant predictive power for high-frequency volatility of a variety of domestic commodity,except that this predictive power changes with the forecast period.In addition,the predictive power of individual international stock market volatility information usually has a relatively weak robustness,and changing the assessment method,rolling window length and forecasting method all result in changes in the predictive power of some international stock market volatility information.Besides,this thesis introduces the Markov regime switching methods with fixed and time-varying state transfer probability,using FTP-MIDAS-RV-X and TVTP-MIDAS-RV-X models to forecast the high-frequency volatility of domestic commodities with international stock market volatility information.The out-of-sample evaluation results show that Markov regime switching models with fixed state transfer probability are sufficient to capture the regime switching characteristics presented in the high-frequency volatility behavior of most domestic commodities and to improve the forecasting accuracy of the high-frequency volatilities of domestic commodities by making full use of volatility information from individual international stock market.However,the Markov regime switching models perform poorly in predicting the short-term volatility of Corn and PTA.The Markov regime switching method also improves the short-term volatility forecasting accuracy of soybean meal(Soybean M),but the improvement is limited.In addition,we design the FTP-MIDAS-LASSO and TVTP-MIDAS-LASSO models by combining Markov regime switching techniques with LASSO method and MIDAS-RV model,and compare their forecasting performance with MIDAS-LASSO model,combination forecasting methods,dimensionality reduction techniques and Markov shrinkage models in the HAR framework.The out-of-sample prediction evaluation results found that the FTPMIDAS-LASSO and TVTP-MIDAS-LASSO models exhibit better prediction performance than the combination forecasting methods,dimensionality reduction techniques and the HAR framework-based Markov shrinkage models.Finally,this thesis discusses the application value of commodity volatility forecasting using the average expected utility calculation method of Bollerslev et al.(2018)and the realized volatility-based value-at-risk measure.The results of the discussion on the application value prove that accurate forecasting of commodity volatility is helpful for investors and risk managers to achieve relatively desirable economic returns and to anticipate the magnitude of their future exposure to losses.