农产品期货市场波动率的动态特征及其预测模型

Realized Volatility Dynamics and Forecasting Model for Agricultural Commodity Futures

本文以8种农产品期货的高频数据为样本,实证考察了我国农产品期货市场已实现波动率的动态特征,发现农产品期货已实现波动率同时具有长记忆性和区制转换性。在此基础上构建了长记忆马尔科夫区制转换模型来预测农产品期货的已实现波动率,并比较和评价了该模型与其他嵌套模型的预测性能。结果发现,我国农产品期货的已实现波动率具有高波动和低波动两种不同的状态,状态之间的转换概率较小,低波动状态的稳定性比高波动状态强;同时引入长记忆性和区制转换能进一步提高模型的预测性能,长记忆马尔科夫区制转换模型是预测性能最好的模型。

This paper empirically analyzes the dynamic properties of the realized volatility in China＇ s agricultural commodity future markets by employing the high - frequency data of eight agricultural commodity futures, and finds that the realized volatility exhibits both long memory and regime switching features. To capture these properties simultaneously, we modeled the realized volatility of agricultural commodity futures in China＇ s future markets by a MS -ARFIMA model, and compared its forecasting performance with that of a variety of competing models which have been encompassed by the proposed model. The estimated results show that the dynamics of realized volatility for agricultural commodity futures are characterized by two levels of long memory： one associated with the low - volatility regime and the other one associated with the high - volatility regime, and the probability to stay in the low - volatility regime is higher than that of the high - volatility regime. The outsample volatility forecast results show that the combination of long memory with switching regimes can improve the out - of - sample realized volatility forecast performance, and the proposed MS - ARFIMA model represents the superior out - of - sample realized volatility forecasts over those obtained using nested models in China＇ s agricultural commodity future markets.