Abstract: The coexistence of high growth and high volatility over recent years in China’s commodity futures market where even “roller coaster” market conditions emerged in parallel with the international bulk commodity market have aroused wide concern over potential risks. Therefore, identifying and measuring commodity futures market risks is worth an in-depth exploration. In this paper, an observation-driven model based on score function is used to construct a dynamic semi-parametric model to estimate the expected shortfall and value at risk to measure and predict the risk of China’s commodity futures market. The results show that the generalized autoregressive scoring models based on student-T distribution, skewed student-T distribution, and asymmetric Student-t distribution have a good application. In this paper, the single-factor generalized autoregressive score model applied to risk measurement shows that the average loss of the single-factor generalized autoregressive score model is relatively smaller than that of the two-factor model, and the risk change of commodity futures can be observed more dynamically. The average Expected Shortfall of chemical futures is rather prominent, but the fluctuation range is small. In contrast, the Expected Shortfall extreme value of energy futures is higher, and the expected loss is as low as -13.03%. Comparatively speaking, the risk degree of the commodity futures market of agricultural and sideline products is relatively low. Commodity futures have exhibited varying degrees of risk accumulation in the near term since January 2021. Energy, chemicals, grains, soft commodities and grease oils futures showed significant expected losses from March 22 to April 14, 2021, while precious metals and non-ferrous metals futures showed significant expected losses from January 12 to January 15, 2021. Under the background of global economic recovery, the effective avoidance of commodity market risks has attracted the attention of market participants, policymakers and researchers. For the risk monitoring of the commodity market, we should pay attention to the impact of exogenous events and the persistence of risk events and carry out dynamic measurement of futures market risks.

Key words: commodities futures, GAS model, ES-VaR, dynamic semi-parametric model, single-factor estimation