摘要我国原油长期面临高需求、低自产的格局。缺乏成熟的原油期货市场导致我
国虽作为全球最大的原油需求方，却迟迟未能掌握国际原油的定价权和话语权，
在原油贸易中常处于被动地位。2018年3月26日，以人民币计价、面向全球投
资者的上海原油期货在上海国际能源交易所正式推出。上海原油期货的适时推出，
将有助于扭转这一不利局面，对我国对冲原油贸易风险敞口、主导国际原油定价
权、推动人民币国际化等战略目标均有重大意义。一直以来，套期保值者都是原
油期货市场的主要参与群体之一。在投资实践中，上海原油期货在对原油现货价
格风险的对冲中是否具有较好的套期保值绩效，是其能否成功吸引投资者，形成
规模市场的决定性因素之一。
学界、业界常用的方差套期保值目标函数只考虑了最小化以方差度量的资产
组合整体风险这一单一目标。与之不同，本文所采用的条件风险价值CoVaR 着
重度量单侧下行风险，考虑了投资者的风险偏好，兼顾了期望收益，在理论上是
更适合实际套期保值应用的目标函数选择。在期货、现货收益率的时间序列计量
建模中，多元广义自回归得分（Multivariate Generalized Autoregressive Score）模
型设定具有在参数估计中对异常值稳健、能充分利用历史分布信息等优点，因而
能为套期保值策略的构建提供更为准确的条件分布信息，有助于提升套期保值投
资实践的绩效。本文推导了学生t 分布假设下的最小CoVaR 套期保值比率，以
及多元学生t 分布下的GAS（MV-t-GAS）设定，将此二者结合应用于上海原油
期货的套期保值中，具有一定的理论价值。
实证结果表明，与具有重要国际地位的Brent 原油期货相比，上海原油期货
在对胜利、沙特和大庆原油现货价格的风险对冲绩效上优势明显，不过在对Brent
原油现货价格的对冲上表现不佳。这表明在不同油品之间、国内国际之间均存在
一定的原油市场价格异质性差异，主要表现为相关系数上的显著差异。就目前而
言，上海原油期货为国内各品种原油的市场参与者以及国际中质含硫原油的套期
保值者提供了更有效的风险对冲工具，有望吸引更多投资者，并主导中质含硫原
油市场以及亚太原油市场定价权。本文采用的GAS 设定下的CoVaR 动态套期保
值模型在样本内、外都能有效降低下行风险和整体风险，具有实践应用价值。随
着投资者风险厌恶程度提升，最小CoVaR 动态比率的波动（标准差）趋于平稳，
并会逐渐接近于最小方差比率。同时，动态套期保值组合的整体风险（方差）控
制绩效会逐渐改善、期望收益率会逐渐降低。
关键词：CoVaR，MV-t-GAS 模型，套期保值，上海原油期货
I
Abstract
Abstract
For a long time,there exists a dilemma of high demand and low self-production
of crude oil in China.As the world's largest crude oil demander,however,China does
not have pricing power and discourse power of international crude oil and is often in a
passive position in crude oil trade due to the absence of mature local crude oil futures
market.On March 26,2018,Shanghai crude oil futures,which is denominated in RMB
and facing global investors,was officially launched in Shanghai International Energy
Exchange.The timely launch of Shanghai crude oil futures will help to reverse the
adverse situation,which is of great significance to China's strategic targets such as
hedging crude oil trade risk exposure,leading international crude oil pricing power and
promoting RMB internationalization.Hedgers have always been one of the main
participants in the crude oil futures market.In actual investment,whether Shanghai
crude oil futures have a satisfactory performance in hedging the fluctuation risk of crude
oil spot price is one of the decisive factors for its success in attracting investors and
forming a large-scale market.
The Variance objective function commonly used in academia and industry only
considers the single goal of reducing the overall (bilateral)risk of hedging portfolio
measured by variance.Differently,the conditional value at risk (CoVaR)used in this
paper,which considers the risk preference of investors and the conditional expected
return,focuses on downside (unilateral)risk rather than bilateral risk of portfolio.Thus,
theoretically,it is an objective function which is more suitable for practical hedging
application.In the time series modelling of spot and future return yields,the burgeoning
Generalized Autoregressive Score (GAS)model specification offers many advantages,
such as outlier robustness,fully usage of historical information et al.in the parameter
estimation process.Thus,it can provide more accurate conditional distribution
information to the construction of hedging portfolio and help to improve the
performance of hedging investment practice.This paper deduces the minimum CoVaR
hedging ratio under the student t-distribution assumption and the GAS specification
under the multivariate student t-distribution (MV-t-GAS),which has a certain
theoretical valuebination of the two is applied to the practical application of
Shanghai crude oil futures hedging strategy.
The results show that compared with Brent crude oil futures which has important
I
Abstract
international status,Shanghai crude oil futures can effectively hedge the price risk of
Shengli,Saudi Arabia and Daqing crude oil spot,but it performs poorly in hedging
Brent crude oil spot.The discovery indicates that there is not only heterogeneity in the
prices of different crude oil products,but also heterogeneity between Chinese market
price and international market price.Main performance of the heterogeneity is the
significant difference in correlation coefficient.At present,Shanghai crude oil futures
is a more effective risk hedging tools for Chinese market participants of various types
of crude oil and international hedgers of medium sour crude oil.Thus,Shanghai crude
oil futures is expected to attract more investors and dominate the pricing power of
medium sour crude oil market and Asia-Pacific crude oil market.In the other hand,the
CoVaR dynamic hedging model under GAS specification used in this paper can
effectively reduce the downside risk and overall risk,no matter in or out of sample,
which proves its value in practical application.With the increase of investor’s risk
aversion degree,the minimum CoVaR dynamic hedging ratio will gradually approach
the minimum variance hedging ratio,and its fluctuation (standard deviation)tends to
be stable.Meanwhile,the dynamic hedging portfolio’s overall risk (variance)control
performance will gradually improve,but its expected return will gradually decrease as
a price.
Keywords:CoVaR,Multivariate-t-GAS model,Hedge,Shanghai crude oil futures
II
目录
目录
摘要............................................................................................................................I
Abstract ..........................................................................................................................I
第1章绪论................................................................................................................1
1.1研究背景.............................................................................................................1
1.2研究意义.............................................................................................................2
1.2.1理论意义.......................................................................................................3
1.2.2现实意义.......................................................................................................3
1.3研究内容与研究方法.........................................................................................4
1.3.1研究内容.......................................................................................................4
1.3.2研究方法.......................................................................................................5
1.4本文的创新与不足.............................................................................................5
1.4.1创新之处.......................................................................................................5
1.4.2不足之处.......................................................................................................6
第2章文献综述........................................................................................................7
2.1套期保值目标函数文献综述.............................................................................7
2.1.1组合风险最小化目标...................................................................................8
2.1.2投资者效用最大化目标...............................................................................9
2.2期货、现货计量建模文献综述.......................................................................13
2.2.1传统静态建模方法.....................................................................................13
2.2.2传统动态建模方法.....................................................................................13
2.2.3广义自回归得分（GAS）建模方法.........................................................14
2.3文献述评...........................................................................................................17
第3章GAS 设定下的CoVaR 动态套期保值模型...............................................20
3.1CoVaR 套期保值模型.......................................................................................20
3.1.1套期保值理论基础.....................................................................................20
3.1.2套期保值目标函数选择.............................................................................21
3.1.2.1传统方差目标函数及其不足..............................................................21
3.1.2.2CoVaR 目标函数的一般形式..............................................................22
3.1.3学生t 分布下的CoVaR 套期保值比率推导............................................23
3.1.4套期保值策略绩效评估指标.....................................................................26
I