套期保值,估计风险与贝叶斯统计——基于中国铜期货市场的经验研究

Futures Hedging,Risk and EstimationBayesian Statistics——Based on Empirical Research In China's Copper Futures Market

针对期货最优套期保值策略估计中可能存在的估计风险问题,本文对单变量线性回归模型(OLS模型)和多变量线性回归模型(VAR模型和EC-VAR模型)进行贝叶斯分析,并采用Gibbs抽样方法对中国铜期货市场的最优套期保值策略进行了实证分析。本文还同时估计了基于频率统计方法的最优套期保值策略,并对贝叶斯统计下和频率统计下的最优套期保值策略进行了分析比较。实证结果清楚表明,估计风险对模型结果有重要影响。在处理估计风险方面,贝叶斯统计较频率统计方法有明显优势。

The risk of econometric models includes model-misspecification risk and estimation risk. Backward-looking econometric models based on frequentist statistics doesn＇t account for the existence of estimation risk. The Bayesian approach provides a general framework where estimation risk is naturally accounted for when considering the parameters as random variable. This article uses Bayesian approach based on MCMC simulation to estimate the optimal hedge ratio of China＇s copper futures market. The performance of the Bayesian hedge ratios is compared to that of alternative frequentist statistics approach. The Bayesian empirical result indicates EC-VAR model performs best and the hedging performance of VAR model significantly surpasses that of simple OLS model. On the contrary, if not accounting for estimation risk, EC-VAR model performs worst and the hedging performance of VAR model doesn＇t significantly surpass that of OLS model.