我国农产品期货市场的风险测度模型及其后验分析

Research on Stylized Facts and Risk Models for Chinese Agricultural Futures Market

近年来,我国农产品期货市场取得快速发展,但有关该市场波动特征和风险状况的研究却非常缺乏。以我国农产品期货市场中的4种代表性价格指数为例,首先对其价格变化统计特征及波动模式进行了全面深入的检验,然后运用严谨系统的后验分析(Backtesting analysis)方法,分别在多头和空头两种头寸状况以及5种不同分位数水平下,实证对比了8种风险测度模型对VaR(Value at Risk)和ES(Excepted shortfall)两种不同风险指标估计的精度差异。研究结果表明,我国农产品期货市场的价格波动不存在显著的杠杆效应(Leverage effect),但却具有明显的有偏(Skewed)和条件厚尾(Conditional fat-tail)特征。另外,在综合考虑了模型估计效率和风险测度精度后,基于有偏学生t分布的普通GARCH模型是一个相对合理的风险测度模型选择。

Agricultural futures market is an important part of modern financial market systems. Almost every government and industry highly values its functions of guiding, hedging and stabilizing markets. Therefore, the market has been growing. The rapid development of China market is the driving force of economic growth. However, little work has been done to detect volatility features and risk characteristics of China market. The main objective of this paper are to （1） build different volatility models （GARCH; APARCH; FIGARCH; FIAPARCH） for the conditional volatility of returns for Chinese agricultural futures indexes, （2） choose the skewed Student-t distribution, which is more appropriate to depict the typical features of financial assets returns, and （3） fully investigate and describe distribution features of Chinese agricultural futures returns. We compute agricultural futures＇ VaR （Value at Risk） in different models and adopt both unconditional coverage testing and conditional coverage testing. The application scope and different VaR models are constructed to offer the best VaR measurement for Chinese agricultural futures market. Because of the asymmetrical characteristics of financial asset return distributions, derivatives with the same underlying asset have different VaRs when taking a long-term or short-term position. Therefore, it is meaningful to investigate the right tail and left tail separately in asymmetric return distribution. In this paper, we will test different models in both long and short positions to elaborate their effectiveness and practicability in Chinese agricultural futures market. Our main findings are summarized as follows： （ 1 ） The return of Chinese agricultural futures market exhibits relative significant Leptokurtic, fat tailed and skewed distributions, and volatility clustering effect. However, there is no evidence that the volatility of agricultural futures market presents leverage effect like Chinese stock market; （2） Using skewed Student-t distribution can help improve the accuracy of VaR and ES estimation in Chinese agricultural futures market. However, adding the leverage effect and long-memory in volatility models does not benefit the accuracy of VaR and ES estimation ; （3） When considering both efficiency of model estimations and accuracy of calculations, we think GARCH-SST model is an excellent choice for risk measurement in Chinese agricultural futures market. This paper has important practical and social implications. We make the case that some statistical characteristics are the stylized facts and GARCH-SST model is a reasonable choice to forecast VaRs of Chinese agricultural futures market. These techniques and empirical results can offer useful theoretic reference and practical approaches for risk measurement in Chinese agricultural futures market. Furthermore, the findings of stylized statistical characteristics of Chinese agricultural futures market have important regulatory sense in position limits and margin ratio, et al.