Bollerslev (1986)提出一般化自我回歸條件異質變異數模型，即GARCH模型。此模型雖然可捕捉波動叢聚(volatility clustering)現象，但Chan (2003)考慮跳躍行為變異數與共變異數，提出雙變量跳躍強度隨時間變動模型(correlated bivariate Poisson with GARCH, CBP-GARCH)，可以估計兩種資產共同跳躍強度隨時間變動的現象，捕捉兩資產價格不連續的變動及兩資產報酬的共變異數關係。 農產品價格波動受天氣、政策、市場供需等影響甚劇，而農產品價格與期貨價格可能皆會因稀少事件的發生而存在干擾的現象。本研究沿用Chan (2003)的CBP-GARCH模型探討2008年5月9日至2016年5月9日中國農產品期貨與現貨市場為主要目的，其次也研究其與基差風險的變化。主要研究大連商品交易所(Dalian Commodity Exchange，簡稱DCE)的農產品期貨與現貨，並對於樣本期間內全球所發生的事件進行子樣本期間農產品期現貨市場風險特性的分析，相關結果將提供市場參與者未來制定投資決策時，能有一完整參考的研究結果。選用交易規模連續5年位居大連商品交易所首位的豆粕期貨進行實證，結果顯示(1)兩報酬率皆存在著高度的波動叢聚現象，而期貨又較於現貨高；(2)現貨與期貨報酬存在共同跳躍行為，而其變動可能受到某些因素的干擾。 This paper uses Chan (2003) CBP-GARCH model to study jump dynamics in the return of agricultural commodity spot and futures. The traditional continuous and smooth models, like Bollerslev (1986) GARCH model, may fail to capture extreme returns volatility. The performance of CBP-GARCH model is better. CBP-GARCH model has high accuracy because it can capture dynamic jump process and jump correlation. CBP-GARCH model shows that the character of volatility clustering of agricultural commodity in China. Moreover, we add basis in our model. We analyze soybean meal spot and futures in our study. Because Dalian Commodity Exchange occupies the one position in the soybean meal futures in China. In our study shows (1) futures causes greater volatility than spot. (2) The two volatility indices have the individual jump and the joint jump behavior.
